{"paper_id":"d8642a76-4e3e-473b-bd4e-5afe7881af93","body_text":"Evidence of latent-lytic replication in EBV-positive Burkitt lymphoma from whole genome and transcriptome sequencing | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Evidence of latent-lytic replication in EBV-positive Burkitt lymphoma from whole genome and transcriptome sequencing Ismail Legason, Adam Burns, Dimitrios Vavoulis, Silvia Halim, and 27 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7473738/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Epstein–Barr virus (EBV) is a ubiquitous human herpesvirus linked to multiple malignancies, but its role in disease remains uncertain. We analysed 410 EBV genomes, including 64 newly sequenced isolates from plasma of East African children with and without Burkitt lymphoma (BL), integrating whole-genome and transcriptomic data. Population genomic analyses revealed marked regional structure with strain-level diversity shaped by both recombination and mutation, particularly within latency and immune-evasion genes ( EBNA-1 , LMP-1 , BCRF1 , BNLF2a ), reflecting selective pressure from host immunity and tumour microenvironment. We identified a BL-associated mutational signature (SBS_EBV3) enriched for T > C and C > T substitutions in G-rich contexts, with limited similarity to known COSMIC signatures (closest match SBS54, cosine similarity 0.74). Transcriptomic profiling demonstrated a mixed latent–lytic expression programme in BL, potentially promoting recombination and mutagenesis. These findings define new features of EBV evolution in BL and highlight opportunities for diagnostics and vaccines targeting both latent and lytic antigens. Biological sciences/Cancer/Tumour virus infections Health sciences/Medical research/Genetics research Epstein-Barr virus latent-lytic replication recombination mutations genetic variation and Burkitt lymphoma Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction The Epstein-Barr virus (EBV), a double-stranded DNA virus of the γ-herpesvirus family, is the most ubiquitous human pathogen, best known for causing infectious mononucleosis 1 . Human infections occur through saliva, with at least 90% of the global population exposed to the virus at some point during their lifetime 1 – 3 . While most infections are asymptomatic, approximately 1% of cases may lead to complications ranging from neoplastic disorders to systemic autoimmune diseases 3 – 6 . Among EBV-associated neoplasms, Burkitt lymphoma (BL) and nasopharyngeal carcinoma (NPC) represent the two well-characterised models 4 . BL, an aggressive B-cell malignancy, is primarily a paediatric malignancy and highly endemic in equatorial Africa and parts of Oceania, with an estimated annual incidence of 0.17 to 6.2 per 100,000 children in some parts of equatorial Africa 7 . In contrast, NPC, an epithelial malignancy, predominantly affects adults and is highly endemic in Southeast Asia and parts of North Africa, with an incidence of 0.3 to 2.1 cases per 100,000 individuals 8 . Although both malignancies share a strong link to EBV, their distinct geographic distributions, patient age, environmental cofactors ( P. falciparum for BL 9 , and dietary or lifestyle factors for NPC 8 ), as well as previously described host-specific genetic factors 8 , 10 , 11 , point towards a complex interplay of all these variables in determining these divergent disease phenotypes. Notably, the role of specific EBV genome characteristics remains largely elusive. The EBV genome is generally regarded as stable, evolving clonally in most tumours 12 , 13 . After B-cell infection, EBV persists latently during clonal proliferation, germinal centre transit and memory B-cell differentiation, with EBNA-1 as the primary antigen tethering episomal genomes to the host chromatin 3 , 14 . Occasional abortive reactivation events, producing a restricted set of lytic proteins without full viral production, have been proposed as a mechanism for long-term persistence 15 , 16 . However, emerging evidence indicates that a full lytic reactivation of the viral genome may be more frequent in BL than previously recognised 17 , 18 , a process that can amplify viral mutations and promote recombination, thereby accelerating viral evolution. Recombination appears to be more common in settings of high EBV co-infection rates, such as endemic regions 19 , and in immunodeficient individuals 20 , generating mosaic genomes through homologous or non-homologous exchange of genetic material between co-infecting strains 21 . While recombination rates vary across herpesviruses 22 , 23 , studies specifically focused on BL remain limited 19 , and it is not yet clear whether recombination or mutation is the dominant force driving EBV diversification in this context. Mutational processes, including point mutations, small deletions, and polymorphisms, introduce substantial genetic variation with profound biological consequences 24 . For example, EBNA-2 sequence variation defines viral subtypes and modulates B-cell transformation 25 , 26 , while variants of LMP-1, BALF2 , and RPMS1 are linked to poor clinical outcomes in NPC-endemic regions 24 , 27 , 28 . Furthermore, intertypic recombinants are frequently identified in certain EBV-associated conditions 20 . Collectively, these observations suggest that both recombination and mutation act as complementary forces shaping EBV genomic diversity, underscoring the need for comparative analyses of their relative contributions to BL pathogenesis. While recombination events can produce large genomic mosaics 21 , mutations frequently show distinctive signatures that reveal their underlying mechanisms 29 . As such, it is vital to understand both the processes driving EBV evolution and the mutational mechanisms behind them, as this knowledge can help identify biomarkers that may be useful for diagnosis or prognostic assessment. In cancer genomics, mutational signature analysis has proven invaluable for identifying footprints of distinct mutagenic processes. For example, smoking, UV-light or enzymatic activities such as AID- and APOBEC-mediated cytidine deamination, oxidative stress, or defective mismatch all cause specific mutational footprints 30 . EBV-infected B cells undergo somatic hypermutations in germinal centres, where AID is naturally expressed and APOBEC activity may also be elevated 31 , 32 , offering a plausible endogenous mechanism of viral mutation. The APOBEC family comprises cytidine deaminases that primarily edit cellular mRNA, but also act on replicating viruses, forming part of the host innate immune response 32 . Applying a mutational signature framework to EBV genomes may therefore reveal whether host-derived pressures, such as germinal centre-associated somatic hypermutation or APOBEC activity, leave a discernible imprint on the viral genome. This approach can provide deeper insight into the interplay between host immunity, viral evolution, and oncogenesis. To explore these hypotheses, we analysed 64 novel EBV genomes isolated from cell-free DNA (cfDNA) samples, along with 346 geographically representative public genomes, all of which were isolates from primary clinical samples. Our cohort included children and young adults with and without BL in Uganda and Tanzania. cfDNA comprises apoptotic and necrotic cell products released into the bloodstream 33 . In neoplastic disease, these products may include tumour fragments, hence the term ‘circulating tumour DNA’. Because the EBV genome is frequently integrated into the host genome 34 , 35 , analysing circulating tumour DNA offers a non-invasive window into the tumour-associated viral genome. The present analysis aimed to understand the relative contribution of recombination and mutation events to EBV genome diversification in BL compared to other EBV-associated conditions by 1) describing differences in the viral genome across geographic regions and disease phenotypes, 2) assessing the recombination frequency of EBV genomes derived from BL compared to other phenotypes, 3) identifying viral variants enriched in BL, 4) characterising mutational signatures associated with BL using the COSMIC single-base substitution (SBS) model, and 5) linking viral genetic alterations, including recombination and mutational patterns to EBV gene expression programmes in BL. Results Study cohort characteristics An overview of the study is provided in the Supplementary Methods (Suppl. Figs. 1-3). Briefly, participants were identified from the Aggressive Infection-Related East African Lymphoma (AI-REAL) study, a prospective case-control study in Uganda and Tanzania. Out of a 310 sample-study cohort, 53, including BL and other lymphoma samples for whom sufficient EBV ctDNA was available, underwent whole genome sequencing. Eleven EBV-positive healthy controls with sufficient EBV DNA levels were included in the final sample cohort, resulting in a total of 64 primary study samples. The median age of the study cohort was 9 years. There were more males than females (64% vs. 36%), and 77% of the study cohort came from Uganda. Clinical symptoms such as weight loss, cachexia, and night sweats were common among cases, with over half of the individuals diagnosed with advanced-stage tumours (Suppl. Table 1). EBV whole genome sequencing We sequenced the 64 EBV whole genomes using a capture-based protocol. The average sequencing depth across the runs was 271.7x in the present study, with 92.6% of the EBV genome covered by at least 10x coverage (Suppl. Fig. 4). A total of 36,963 variants (SNVs and indels < 50 bp) were identified across the 64 genomes. Per sample average, there were 641 variants in BL compared to 625 in other malignancies and 290 in EBV-positive healthy controls. To verify the accuracy of sequencing and variant calling, we processed samples in duplicate and called variants to assess the reproducibility. The results showed that the coefficient of variation (CV) between the first and second run was less than 4% (Suppl. Table 2). Summary of the published genomes Next, we combined data from these new samples with published data to create a dataset of 410 EBV whole genomes, comprising 332 samples from individuals with malignancies and 78 from EBV-positive healthy controls, all derived from EBV DNA extracted either from primary tumour samples or plasma (for information on analysis samples, see Supplementary Data 1). Most of the public EBV genomes lacked strain information. Using the Basic Local Alignment Search Tool (BLAST) approach, we were able to identify 266 (76.9%) of these genomes as Type 1 and 80 (23.1%) as Type 2 (Suppl. Fig. 5). Mixed or undetermined strains were excluded from further analysis. This analysis was vital in ensuring we ordered the contigs obtained from the de novo assemblies against the correct EBV reference genome. The mean sequencing depth for the public data set ranged from 141.8x to 776.92x, and genome coverage varied between 75.0% and 98.6% across studies. We prioritised samples with at least 10x depth and 75% genome coverage for our downstream analysis (Suppl. Fig. 4). EBV genomes display population structure shaped by region, with no evident phenotypic clusters We analysed EBV genome diversity using phylogenetic and dimensionality reduction techniques. A total of 410 EBV genomes, including reference genomes (NC_007605.1 for EBV Type 1 and NC_009334.1 for Type 2) and two well-characterised EBV-positive B cell lines, Daudi and Raji, were used to build the phylogenetic tree. These genomes originate from five regions: Africa (Kenya, Malawi, Uganda), Asia (China), Central America (Guatemala, Peru), North America, and South America (Suppl. Fig. 3) The phylogenetic results revealed a clear geographic population structure of the EBV genomes, clustering by region regardless of the disease phenotypes involved (Fig.1). Within African samples, we observed no distinct patterns between cases and EBV-positive healthy controls. Similar patterns were also observed in Asian samples, with no evident phenotypic clustering. Out of 141 BL samples, two were from Asia, and these formed a subcluster with Asian NPCs; the African BL cases clustered with healthy African samples. In terms of strains, the larger proportion of African type 2 EBV genomes formed a distinct subcluster from type 1 genomes, irrespective of the disease phenotype. Interestingly, four of the NPC-associated EBV genomes from Asia clustered with type 2 African EBV genomes. Two of these were previously reported as Type 1, and the other two were not assigned any genotype in the original study 24 . However, when we performed alignment (BLAST) against EBNA2 -type-specific contigs, these genomes showed the strongest hits for type 2 EBV. Hence, we categorised them into type 2 genomes. The heatmaps: the outer ring represents sample phenotypes, the middle ring denotes the country of origin, and the inner ring indicates the region. The phenotypes analysed in this study include BL, EBV-positive healthy individuals, HL, NKTCL, NPC, and other rare EBV-associated malignancies. The tips are annotated with EBV type (type 1 green triangle; type 2 blue dot), and the most recent ancestor (tree root) is indicated. The support values shown at the nodes represent TBE values computed from RAxML-ng; nodes with support values greater than 70% are displayed. To corroborate our phylogenetic results, we conducted principal component analysis on the 410 genomes. A total of 36,100 variants were loaded from the binary PLINK file. Thirty-four thousand three hundred twenty-nine (34,329) variants were excluded due to minor allele thresholds (VAF < 0.05), and an additional 995 were pruned due to linkage disequilibrium (r² > 0.6), resulting in a final set of 776 variants used to generate the relationship matrix. In the PCA, the origin of samples or genomes contributed most to the genetic variation (Fig. 2). The first two principal components (PC1 & PC2), which explained ~ 34% of the total variance, showed a strong clustering by geography (Africa vs. Asia vs. the Americas). PC1 separated African strains from Asian strains, accounting for 18.8% of the variation. PC2 explained 15.4% of the genetic variation and distinguished American EBV strains from those of Asian and African origins. Consistent with the phylogeny, no apparent phenotypic clustering was observed. EBV strains (Type 1 vs Type 2) were dispersed across regions and did not appear to define principal axes of variation. Even in higher-order PCs (PC3-PC6) (Suppl. Fig. 6), the clustering was weak by strain, suggesting that intra-strain variation is minor compared to regional diversity. The scatter plot illustrates the genetic variation in EBV genomes. The first two principal components account for 18.8% and 15.4% of the variation, respectively. Points are coloured by region, with corresponding phenotype annotations shown as a heatmap below the scatter plot. Additional tracks display normalised PC1 and PC2 values (z-scores), emphasising systematic differences between regions and related disease phenotypes. Samples from Africa (red) cluster closely on the right side (high PC1). Samples from Asia (green) cluster on the lower left (low PC1, low PC2). American samples (Central/North/South) exhibit greater dispersion, primarily to the left on PC1, yet remain high on PC2, thereby distinguishing Asian strains from American strains. Increased EBV recombination in BL compared to EBV-positive healthy controls and other EBV-associated malignancies, as suggested by linkage disequilibrium analysis To explore differences in the degree of recombination events contributing to the clonal evolution of EBV in disease phenotypes, we calculated pairwise linkage disequilibrium (LD) correlation coefficients (r²) between genomic sites within a 10 kb sliding window for up to 99,999 single-nucleotide variants. Our results revealed steeper LD decay patterns for BL (Fig. 3), NPC, and EBV-positive healthy controls, with mean r² ranging from 0.3 to 0.45 over genomic distances of ~500 bp to 5000 bp, compared to HL, NKTCL, and other EBV-associated malignancies, which maintained higher mean r² values over longer distances before decaying. At longer distances, e.g., > 5000 bp, the mean r² values generally plateaued (mean r² ~ 0.3). At the longest distances (~10,000 bp), most groups converged to low mean r² values. Impact of recombination on EBV genome diversification in BL compared to EBV-positive healthy controls Given the strong recombination signals observed in BL-derived EBV genomes, we investigated whether these events were uniformly distributed across the viral genome or concentrated in specific regions, and how they contributed to viral diversification compared with EBV-positive healthy carriers. We analysed 141 BL cases and 78 EBV-positive healthy controls. Whole-genome alignments (trimmed for poorly aligned regions and gaps) were screened using RDP4 36 , identifying putative recombination in most isolates, with 43 events reaching statistical significance (Suppl. Table 3). To assess their impact, we reconstructed phylogenies of the 219 EBV genomes with and without recombinant regions. The topologies differed markedly, with a normalised Robinson–Foulds distance of 0.884, confirming strong recombination-driven incongruence (Extended Data Fig. 1). We next quantified the relative contribution of recombination to EBV genome diversification–accounting for mutations using ClonalFrameML 23 . Gene-level estimates showed substantial heterogeneity, with wide 95% highest posterior density intervals (HPDIs; Fig. 4). Two latency genes, LMP-1 and EBNA-1 , displayed recombination rates well above the genome average in BL, whereas the tegument gene BOLF1 was enriched for recombination in healthy carriers. Other genes had rates near background levels with overlapping HPDIs and no statistical significance. Each point represents the posterior mean of ρ per base pair (ρ/bp) for a gene, as estimated by ClonalFrameML. Top panel: EBV genes from BL. Bottom panel: EBV genes from EBV-positive healthy controls. Genes are ordered by their recombination rate. Vertical lines indicate the 95% highest posterior density intervals (HPDIs). The shaded grey area in each panel represents the genome-wide 95% HPDI for ρ/bp. Genes are colour-coded across panels, with each gene assigned a unique colour to highlight concordance or divergence in recombination across host groups. Statistical significance is considered when the gene’s HPDI interval does not overlap with the genome-wide 95% HPDI (grey area). The gene-specific error bars show intragenic heterogeneity in the recombination rate. The large 95% HPDI reflects more variability in the ρ/bp signal. EBV latency and immune evasion genes are significantly mutated in BL compared to other EBV-associated phenotypes To assess the impact of mutations on EBV genome diversity and the risk of endemic BL, we quantified the frequency of single-nucleotide variants (SNVs) and indels across the viral genome. A total of 410 EBV genomes from BL, HL, NPC, NKTCL, EBV-positive healthy controls, and other EBV-associated malignancies were analysed. Our results showed that the per-sample mutation frequency of the EBV genome was similar in BL and EBV-positive healthy controls (4.3 per kb), but higher in NKTCL (7.98 per kb), NPC (5.3 per kb), HL (4.8 per kb), and other cancers (4.4 per kb), respectively (Suppl. Fig.7). However, genome-wide analysis of EBV single nucleotide variants (SNV) in BL samples (141 cases) revealed significant regions of sequence variation (hypervariable) and conservation (Fig. 5). The hypervariable regions corresponded to key immune evasion genes such as LMP1 –oncogenic signalling 37 , EBNA-1 , BNLF2a (TAP inhibitor 38 ), BILF2 ( viral G protein-coupled receptor 39 ), and BCRF1 (viral interleukin-10 40 ), as well as lytic genes BBLF4 (component of viral helicase primase ), BLRF1 (gN), BBRF3 (gM), BGLF3 and RPMS1 (BART transcripts) 41 . The conserved regions corresponded to the EBNA-1 Gly-Ala, LF3 (IR4) , and the LMP2A terminal repeats (TR) of the viral genome. The line plot shows the mean SNV count per 1 kb window across the EBV genome in BL. Red dots indicate hypervariable regions (significantly elevated SNV density), while blue dots represent SNV deserts (conserved regions). Genes corresponding to hypervariable and conserved regions are labelled. The gene track was annotated using the B95.8 (NC_007605.1) reference genome. Directional arrows represent EBV genes, coloured by functional group (Latency, Early, Immediate-Early, Late, Immune Evasion, Uncharacterized). We next performed gene-level mutation enrichment analysis across six EBV-associated phenotypes (BL, NPC, HL, NKTCL, other cancers, and EBV-positive healthy controls). Mutation fold enrichment was calculated as log₂(observed/expected) per gene (Supplementary Data 2). Latency genes showed the strongest enrichment in lymphoma phenotypes (except HL), with BL exhibiting the broadest and highest fold changes, particularly in EBNA-1, EBNA-LP, EBNA-3A, LMP-1, and LMP-2A/2B (e.g., LMP-1: log₂FE = 7.5 in BL vs. 6.7 in NPC and 5.9 in healthy controls; Extended Data Fig. 2). In contrast, structural genes (e.g., BLLF1, BLLF2, BDLF3, BPLF1, BRRF2, BNRF1, BGLF5, BGLF1) showed limited or sporadic enrichment. When grouped by functional categories (capsid, envelope, immune evasion, latency, replication, tegument), pairwise comparisons between BL and healthy controls revealed no significant differences (Suppl. Fig. 8). Cliff’s delta indicated small-to-moderate effect sizes, with the strongest difference in immune-evasion genes (Δ = 0.556). To complement this gene-level analysis, we performed a genome-wide association study (GWAS) of 933 polymorphic loci in 141 BL cases and 78 EBV-positive healthy controls. Logistic regression adjusted for viral population structure (PC1, PC2), geography, and strain revealed no variants reaching genome-wide (p < 5 × 10⁻⁸) or suggestive (p < 1 × 10⁻⁵) significance. Several loci showed nominal associations (p < 0.05) but did not warrant fine mapping (Suppl. Fig. 9, Table 4). Enrichment of G-rich mutation signatures in EBV genomes from endemic Burkitt lymphoma Having established that EBV genomes from BL samples exhibited a higher mutational burden compared to EBV-positive healthy controls, we next sought to investigate the mechanisms underlying this diversification. To this end, we applied non-negative matrix factorisation (NMF) to deconvolute the mutational processes shaping the EBV genome diversification. This analysis identified five distinct mutational signatures (SBS_EBV1–SBS_EBV5) across BL, EBV-positive healthy controls, Hodgkin lymphoma (HL), NK/T-cell lymphoma (NKTCL), nasopharyngeal carcinoma (NPC), and other EBV-associated malignancies (Extended Data Fig. 3). SBS_EBV1 showed strong enrichment of C > T transitions, particularly at C-rich trinucleotide contexts, e.g., C[C>T]C, T[C>T]C, C[C>T]A and C[C>T]T. SBS_EBV2 showed dominance in C>T and T>C transitions with a bias towards G- and C-rich flanks, e.g., G[C>T]C, C[T>C]G, G[T>C]C, G[C>T]G). SBS_EBV3 was dominated by T>C and C>T transitions in G-rich contexts, with peak contributions at C[T>C]G, G[T>C]G, and A[T>C]G. SBS_EBV4 showed a strong bias for C>T transitions at CpG sites, while SBS_EBV5 was enriched in C>A transversions at CpC, CpA, and CpG contexts. Phenotype-specific enrichment analysis revealed notable differences (Fig. 6): SBS_EBV1 was significantly enriched in NKTCL; SBS_EBV2 was largely absent from disease phenotypes but detectable in EBV-positive healthy carriers; SBS_EBV3 was significantly enriched in BL relative to all other groups; SBS_EBV4 showed moderate activity in HL and other rare malignancies but was near background in BL, NPC, and NKTCL; and SBS_EBV5 was broadly active across malignancies. To relate these EBV signatures (SBS_EBV1–5) to known mutational processes in human cancers, we calculated the cosine similarities to the COSMIC SBS v3.2 reference set (Fig. 7). SBS_EBV1 showed the highest similarity with SBS30 (cosine similarity = 0.90), SBS_EBV4 and SBS_EBV5 were highly similar to SBS1 and SBS45 (cosine similarity > 0.9), respectively. By contrast, SBS_EBV2 and SBS_EBV3 demonstrated only modest similarity to known signatures, with SBS42 (cosine similarity = 0.65) and SBS54 (cosine similarity = 0.74) as their closest matches. SBS_EBV3 constituted a distinct mutational spectrum in BL samples, suggesting a BL-specific mutagenic process. No EBV signatures matched the canonical APOBEC-associated COSMIC signatures SBS2 and SBS13, which are characterised by predominance of C > T and C > G substitutions in TpCpW trinucleotide contexts 30 . The plot displays effect size estimates (β coefficients) from beta regression models assessing the association between EBV mutational signatures and phenotypes. Bars show the direction and magnitude of effects relative to the reference group (healthy controls), with positive values indicating increased signature activity and negative values indicating decreased activity. Effect sizes (β coefficients) with 95% confidence intervals are shown. Phenotypes analysed included Burkitt lymphoma (BL), EBV-positive healthy controls, Hodgkin lymphoma (HL), NKTCL, nasopharyngeal carcinoma (NPC) and other EBV-associated conditions. The bar plots show cosine similarity between each EBV-derived mutational signature (SBS_EBV1 to SBS_EBV5) and the top five matches to COSMIC single-base substitution (SBS) signatures. The x-axis represents the cosine similarities, and the y-axis represents the Cosmic SBS signatures. The highest similarity was observed for SBS_EBV1, SBS_EBV4, and SBS_EBV5, which closely matched COSMIC SBS30, SBS1, and SBS45 with a cosine similarity greater than 0.9. While SBS_EBV2 and SBS_EBV3 showed weak matches to COSMIC SBS 42 and 54, suggesting distinct mutational signatures in EBV-associated malignancies. Heterogeneous latent-lytic reactivation and replication events of EBV in endemic Burkitt lymphoma Building upon our observations of an increased recombination rate, significant numbers of mutations in latency and immune evasion genes, and the enrichment of a distinct mutational signature (SBS_EBV3) in BL-associated EBV genomes, we sought to understand how these reflected on the viral transcriptional landscape. The use of primary tumour tissues from 15 additional patients for this analysis also allowed us to independently assess our previous findings from circulating tumour DNA sequencing. In addition to the noncoding EBV transcripts ( EBER-1 and EBER-2 ), EBV genomes in BL displayed relatively higher expression levels of both latent and lytic genes in most tumour samples analysed (Suppl. Fig. 14). Thirteen out of fifteen tumour samples (86.7%) showed detectable expression of two or more latency-associated genes. EBNA-1 , characteristic of the latency I programme in BL, was detected in 9/15 (60%) of the samples. Latency II-associated genes LMP-1 , LMP-2A or B were expressed in 12/15 (80%) of tumours, while latency III-associated genes EBNA-2 , EBNA-3A , EBNA-3B and EBNA-3C were detectable in over 60% (9/15) of the samples. Notably, 80% (12/15) of the tumours exhibited detectable expression of lytic genes alongside latent genes, suggesting that full lytic replication occurs in most of the BL tumours analysed. The lytic switch gene BZLF1 was detectable in 10 out of 15 (66.7%) of the samples. Notably, most of the latent and immune evasion EBV genes ( LMP-1, EBNA-1, EBNA-2, EBNA-3A, EBNA-3B, EBNA-3C, EBNA-LP, BCRF1, BILF2, and BZLF1 ) and lytic genes that we identified as significantly affected by mutations were actively transcribed in BL tumours (Fig. 8) . Overall, these results suggest a non-canonical EBV latency programme in BL, with most viral episomes initiating lytic reactivation, as indicated by BZLF1 expression in the majority of tumour samples analysed. The plot illustrates the expression levels of EBV genes identified as mutated in the study. Normalised counts of the EBV genes are shown for 15 BL tumours analysed through FFPE ribosome-depleted total RNA sequencing. The grey violin and boxplot display the expression levels of 10 selected human housekeeping genes (ACTB, GAPDH, HPRT1, B2M, PP1A, RPLP0, RPL13A, TBP, GUSB, and SDHA). The red dashed line marks a normalised count of 1. Latent and immune evasion genes are coloured, while lytic genes remain uncoloured to aid comparison. Discussion The Epstein-Barr virus (EBV) is one of the most widespread human pathogens, typically persisting as a lifelong latent infection 42 – 44 . Although EBV genomes in tumours were once regarded as clonally evolving 12 , 13 , our analysis of 410 genomes, including 64 newly sequenced genomes from BL patients and EBV-positive healthy individuals in East Africa, reveals a more complex and dynamic evolutionary landscape. Consistent with previous findings 45 – 47 , the EBV population was shaped primarily by geography rather than disease phenotype, suggesting founder effects, host-driven selection, and recurrent population bottlenecks. Within the African cohorts, both type 1 and type 2 EBV clusters were observed, though regional signatures dominated over viral subtype, underscoring the influence of host-virus co-evolution. Recombination emerged as a dominant force in EBV evolution, consistent with previous research 19 , 23 . Linkage disequilibrium (LD) analysis revealed elevated recombination rates in BL compared to healthy carriers and other EBV-associated malignancies. Conversely, LD decay was slower in NKTCL and HL, suggesting reduced recombination 48 . Phylogenetic incongruence and genome-wide signals further supported recombination as a pervasive driver of EBV evolution, particularly enriched in immune evasion and latency genes such as EBNA-1 and LMP-1 in BL cases, highlighting the role of host immune pressure and tumour microenvironment in selecting recombinants with enhanced fitness 37 , 38 . In parallel, EBV genomes displayed a heterogeneous mutational landscape, with hotspots in latency-associated and immune evasion genes (e.g., LMP-1 37 , BNLFa 40 , BILF2 39 , and BCRF1 40 ) , but conservation in regions essential for viral genome maintenance ( EBNA-1 Gly-Ala repeat 49 , LF3 -IR 50 , and LMP-2A 51 ). The frequent recombination rates and mutational burden in immune-modulatory genes suggest an active role for viral diversification in tumour immune escape 38 , 40 . Mutational signature analysis identified a novel BL-enriched profile (SBS_EBV3), distinct from known COSMIC single-base substitution (SBS) signatures. SBS_EBV3, characterised by T > C and C > T substitutions in G-rich sequence contexts, may reflect oxidative DNA damage, potentially amplified by EBV-induced reactive oxygen species (ROS) via LMP-1 activity 52 and cytidine deamination mediated by AID overexpression 53 . This signature may highlight a previously unrecognised EBV-specific mutational process contributing to BL pathogenesis. In contrast, SBS-EBV2 was absent from tumours but present in healthy carriers, raising the possibility of a protective effect. Other EBV signatures resembled known COSMIC SBS signatures: SBS_EBV1 for SBS30 (base excision repair defects) 54 , SBS_EBV4 for SBS1 (spontaneous 5-methylcytosine deamination) 55 , and SBS_EBV5 for SBS45–aetiology uncertain but linked to sequencing artefacts in one study 56 and another associated it with better outcomes in urological carcinomas 57 . Importantly, no APOBEC-associated signatures (SBS2 and SBS13) 30 were detected, suggesting that APOBEC editing is not a dominant force in EBV evolution in these cohorts; however, subtle contributions cannot be excluded without motif and strand-bias-aware analyses 58 , 59 . Transcriptional profiling showed that BL-associated EBV genomes did not strictly follow the traditional latency I programme long associated with the disease 3 , 15 . Instead, they exhibited diverse expression patterns of latent and lytic genes, suggesting mixed or ‘intermediate’ transcriptional states. This transcriptional plasticity may facilitate recombination and the accumulation of mutations, aligning with the enrichment of mutations in immune evasion genes and the emergence of SBS_EBV3. This dynamic viral activity challenges the classical binary model of EBV latency in BL. Instead, it supports the idea that latency exists along a functional continuum influenced by complex viral-host interactions in tumour-specific contexts. This study has several limitations that should be considered when interpreting our EBV genome and transcriptome findings. First, although we prioritised high-quality FASTQ data, sequencing artefacts and platform-specific differences could have introduced spurious mutational signatures. We mitigated this through stringent filtering and by requiring consensus across two variant callers. Second, the absence of host genotype data prevented us from controlling for host genetic influences on viral evolution or disease susceptibility, limiting our ability to fully disentangle viral from host–viral interaction effects. Third, the lack of patient-level covariates (e.g., age, sex) in public datasets further limited our ability to adjust for potential confounders. Fourth, the cross-sectional design using archival samples precludes causal inference; we cannot determine whether observed viral differences precede disease or result from tumour evolution and selection pressures. Geographic variation in EBV diversity was partly addressed by including multiple regions and adjusting for viral population structure, but the absence of paired host ancestry data limited our ability to fully correct for population stratification. Finally, transcriptome profiling from FFPE material is inherently constrained by RNA degradation and chemical modification, which may bias coverage and detection sensitivity. We minimised these effects by applying RNA integrity/DV200 thresholds for input selection, using FFPE-optimised library preparation, and applying rigorous downstream normalisation, quality filtering and batch correction. Despite these limitations, this study provides the most comprehensive analysis to date of EBV genomic variation in BL. By integrating phylogenetic, mutational, and transcriptional data, we show that EBV in BL evolves through a combination of recombination, mutation, and transcriptional heterogeneity, yielding distinct genomic signatures not observed in other EBV-associated malignancies. These findings expand our understanding of EBV biology and open new avenues for diagnostics and interventions, including vaccines targeting both latent and lytic viral antigens in EBV-associated Burkitt lymphoma. Future studies integrating host-virus data with functional validation will be essential to establish the diagnostic and prognostic significance of the novel EBV mutational signatures. Materials and Methods Ethics The study received approval from the Oxford Tropical Research Ethics Committee (OxTREC ref: 15-19) in the UK, the National Institute of Medical Research (NIMR/HQ/R.8a/Vol.IX/3408) in Tanzania, the National Council for Science and Technology (HS529ES) in Uganda, and the St Mary’s Hospital Lacor Institutional Research Ethics Committee in Gulu, Uganda. Participants provided informed written consent, and/or assent for minors aged between 7 and 17 years. All study protocols adhered to the Declaration of Helsinki and international data protection regulations. Study participants and Samples Participants in the current study were enrolled in two phases. Phase I (from 2020 to 2024) involved a hospital-based prospective case-control study, originally designed to evaluate the clinical utility of liquid biopsies for diagnosing EBV-driven lymphomas, known as the AI-REAL study. The cases, which included suspected lymphoma patients, were identified from the outpatient and inpatient departments of four tertiary oncology centres in East Africa: St Mary’s Hospital Lacor in Uganda, Muhimbili National Hospital in Dar es Salaam, Kilimanjaro Christian Medical Centre in Moshi, and Bugando Medical Centre in Mwanza, Tanzania. Patients who consented were enrolled and investigated for Burkitt lymphoma (BL) using a liquid biopsy in conjunction with conventional histopathology. The aim was to compare the accuracy of liquid biopsy with traditional histopathology. The liquid biopsy test panel included well-characterised lymphoma targets, markers, and EBV genes. BL patients with sufficient EBV levels (≥ 2 copies per cell) were selected for whole-genome sequencing of EBV. In Phase II (2024), we conducted a population-based survey of children matched to the BL cases by age, sex, and geography. Our field teams were deployed in five BL-burdened districts in northern Uganda, specifically in lower health facilities close to the villages where the BL cases originated and recruited two controls per historical BL case. Peripheral venous blood samples were collected in Qiagen Paxgene blood ccfDNA tubes (Cat no. 768165) containing a preservative that stabilises blood cells and prevents cellular DNA contamination of the plasma. The samples were transported in secure cold boxes maintained at temperatures between 2 °C and 8 °C to St. Mary’s Hospital, Lacor laboratory. At the hospital, samples were separated, initially at a lower speed of 1600 × g for 10 minutes, and the resulting plasma was then centrifuged at a higher speed (4500 × g) for an additional 10 minutes. The double-spun plasma was frozen at -80 degrees until it was required for cfDNA extraction. Informed consent and/or assent from minors aged between 7 and 18 years was obtained from each participant prior to undertaking any study-related procedures. Both studies were approved by the hospital's institutional research ethics committee and the national research regulator, UNCST. Sample processing and EBV whole genome sequencing Plasma samples were thawed, and 4-5 mL was used to extract cell-free DNA with the QIAamp Circulating Nucleic Acid Kit (Qiagen). Samples from healthy children were initially evaluated for EBV using quantitative polymerase chain reaction (qPCR). The kit (catalogue no. A58429) from Thermo Fisher Scientific amplifies the EBNA1 region of the EBV genome via TaqMan probes with fluorophore-based detection. EBV copies per mL were calculated from the Ct values. A total of 78 healthy control samples were assessed, but only 13 showed EBV copies sufficient for downstream analysis (copies > 1000 per mL). These were selected for whole-genome sequencing. DNA libraries were prepared using the Thruplex Tag-Seq kit (Takara Bio), which involved repairing the DNA ends, ligating adapters, and amplifying the library through 7-9 PCR cycles, depending on the input concentration. The resulting libraries were purified and normalised prior to EBV capture and enrichment using the IDT xGen Hybridisation protocol (IDT). The EBV probes were designed with a 60-bp overlap, spanning 120 bp to cover the entire EBV genome. The final library was purified and normalised prior to sequencing on the Illumina MiSeq platform. Selection of public EBV genomes Paired-end FASTQ files from primary studies reporting EBV whole genome sequencing were retrieved from the Sequence Read Archive (SRA) under BioProject IDs PRJNA552587, PRJNA522388 and PRJNA1063319, respectively. A total of 520 paired-end FASTQ files were downloaded and filtered, with only clinical isolates considered for further analysis. Variant calling FASTQs from this study and those downloaded were processed collectively using a custom bioinformatics pipeline. This process involved trimming adapter sequences and removing low-quality bases (Phred score > 20), followed by aligning the reads to the EBV reference genomes (NC_007605.1 for type 1, NC_009334.1 for type 2), as well as a custom hybrid reference (NC_007605.1 combined with EBNA2, EBNA3s contigs from NC_009334.1) using the BWA-MEM alignment method 60 . The aligned reads were sorted and indexed for further analysis. Variant calling was conducted on the aligned BAM files with three different tools: VarScan 61 , VarDict 62 , and Mutect2 63 . Functional annotation was carried out with SNPEff version 5.1d 64 , using annotation databases for both type 1 (NC_007605.1) and type 2 (NC_009334.1) reference genomes. The final VCF files listed variants identified by at least two callers and filtered for a mapping quality score of 60 or higher. Genome assembly Genome assembly was performed using SPAdes v4.2.0, which constructs a de Bruijn graph by decomposing sequencing reads into k-mers of multiple sizes to assemble contigs 65 . In our dataset, the average read length was 143 bp, and the default k-mer sizes used by SPAdes were 21, 33, and 55. The assembly quality was assessed with QUAST v5.3.0, a genome quality evaluation tool 66 . A composite score was calculated based on four QUAST parameters: genome fraction (GF%) > 70%, misassemblies < 1, mismatches per 100 kb < 1000, and GC content within 2% of the reference, as well as a duplication ratio ≤1. Contigs achieving an overall score of 50% or higher were retained for further analysis. To order contigs against the relevant reference genome, a BLAST database was created using EBNA-2 EBV types 1 and 2 contigs, and the contigs were aligned to each reference using blastn, part of the NCBI BLAST+suite 67 . Top matching hits were assigned a strain call, and the resulting EBNA 2 matches were used to group the contigs into types 1 and 2. This categorisation was crucial for the public genomes, which lacked sample metadata and EBV strain information. Subsequently, the assembled contigs were ordered and oriented according to the reference genome using ABACAS (Algorithm-Based Automatic Contiguation of Assembled Sequences) 68 , resulting in a single pseudomolecule FASTA sequence. The assembly quality, including genome coverage, alignment rate, mean depth, and average base quality of aligned reads, was evaluated with minimap2 v2.29-r1283 69 . Contigs were trimmed to eliminate leading and trailing ambiguous bases, along with other sequencing artefacts, using trimAl v1.4 rev 15 70 . Only assemblies with at least 60% coverage of the EBV reference genome and a mean mapping quality of at least 30 were selected for phylogenetic analysis. Evaluating EBV genome diversity and population structure To infer the evolutionary and epidemiological patterns of EBV across various human populations and disease phenotypes, we aligned the FASTA genome assemblies from this study with public genomes using the multiple sequence alignment tool MAFFT v7.520 71 . A total of 410 genomes were aligned, including the EBV reference and the outgroup, Macacine herpesvirus 4 (NC_006146.1). The multiple sequence alignment file was trimmed using trimAl 70 , and known EBV repeat regions were masked using RepeatMasker version 4.1.9 in default mode 72 . The maximum likelihood of the phylogenetic tree was then inferred using the Randomised Accelerated Maximum Likelihood (RAxML-NG v.1.2.2) tool with the General Time Reversible (GTR-G) model 73 . We assessed branch support using the Transfer Bootstrap Expectation (TBE) approach with 50 replicates. Following Lemoine et al. 74 , we considered nodes with TBE > 70% to have strong support. The phylogenetic tree was visualised and annotated using R packages, ape for tree import and manipulation 75 , including re-rooting using the Macacine herpesvirus-4 as an outgroup, ggtree for tree visualisation 76 , ggtreeExtra for adding metadata annotations 77 , and tidyverse for data wrangling and factor reordering 78 . For the linkage disequilibrium (LD) and principal component analysis (PCA), we analysed variants from 410 clinical samples. The variant caller files were merged and filtered for poor-quality variants (MAPQ < 60) and read depth (DP) < 10x. We then split the multiallelic sites into biallelic sites, left-aligned the indels using bcftools 79 and converted the file into PLINK binary format with genotype calls. Prior to linkage disequilibrium pruning, we estimated the average variant density per kilobase of the EBV genome across all samples and used this estimate to set the pruning window size. Variants were filtered by allele frequency (minor genotype frequency > 5%) and LD. We applied a less stringent LD pruning threshold (r² > 0.6, within a 1000-bp sliding window) to retain a sufficient number of informative variants for PCA. A stricter threshold (e.g., r² < 0.2) would remove many variants and risk underrepresenting population structure in compact viral genomes like EBV ~ 172 kb. In total, 776 variants were included in the PCA. All analysis was conducted in PLINK v1.9 80 , a whole-genome analysis toolset. The PCA was visualised in R using tidyverse, patchwork, scale and RColorBrewer packages 81 . Recombination analysis via Linkage disequilibrium First, we estimated recombination frequency in EBV genomes derived from BL and compared it to other phenotypes using linkage disequilibrium (LD) analysis. LD is a standard method to infer recombination and has been widely applied to viral genomes 48,82 . A steeper LD decay indicates frequent recombination events, while a slower decay reflects low recombination, preserving genetic associations over longer distances 48 . We extracted phenotype group-specific single-nucleotide variants (SNVs) filtered by minor allele frequency (MAF > 0.01) and calculated pairwise LD (r² > 0.1) within 10-kb sliding windows. The choice of a stricter LD threshold (r² > 0.1) was to retain only weakly correlated variants and remove strongly linked sites. This approach has been shown to reduce redundant sites and improve the resolution of LD decay patterns, making subtle recombination signals detectable that strong LD would otherwise mask 83 . LD decay (r²) with genomic distance was modelled using non-linear least squares (NLS), and trends were visualised with LOESS smoothing and 95% confidence intervals. Pairwise correlations were computed using PLINK v1.9 80 , with statistical analyses and plotting performed in R 81 . Verification of recombination and gene-level quantification studies As a sanity check, we used the rapid recombination detection programme (RDP4) 36 . This toolkit employs an ensemble of methods: phylogeny-based (RDP and Bootscan) and statistical models (GENECONV, MaxChi, Chimaera, 3Seq, and SiSscan) to identify recombination events and breakpoints in DNA or protein sequences. EBV genomes from 141 BL cases and 78 EBV-positive healthy controls were analysed. Multiple sequence alignments were generated using MAFFT 71 and trimmed using the Gblocks software 84 before loading into the RDP4 programme, with a default window and step sizes of 200 and 20, respectively. The output from RDP4, which includes recombination events, breakpoint positions, recombinant sequences, and major and minor parental sequences, was exported in CSV format for further analysis (Supplemental Table 2). Putative recombination events were only considered if at least five methods, including at least one phylogenetic-based method, detected the event with a threshold p-value < 0.05 after Bonferroni correction. The phylogenetic trees, with and without recombinant regions, were inferred using the maximum likelihood (ML) method implemented in RAxML, employing the GTR model and 1000 bootstrap replicates 73 . The ML tree topologies were compared using the normalised Robinson-Foulds (RF) 85 implemented in the phangorn package in R 86 . The RF metric quantifies the proportion of differing bipartitions between two trees with values ranging from 0 (identical trees) to 1 (completely distinct topologies). To assess the impact of recombination on EBV protein-coding sequences, we extracted gene-specific coding sequences (CDS) from the multiple sequence alignment files of the EBV genomes using a custom Python pipeline. The alignment file was read with Biopython (v1.85) 87 , and the reference genome (NC_007605.1) was used to map alignment coordinates to genomic positions. Gene coordinates from a BED annotation file were grouped by gene, and exons were merged to generate contiguous CDS sequences, accounting for wraparound genes ( LMP-2A ). For each gene, concatenated CDS alignments were exported as FASTA files using pandas (v2.3.0) 88 for BED parsing and defaultdict structures for sequence assembly. Gene-based phylogenetic trees were inferred independently for each alignment using IQ-TREE v3.0.1 89 with the default GTR+G substitution model and bootstrapping. The trees and FASTA alignments were then used to estimate recombination parameters in ClonalFrameML v1.13 90 . The choice of ClonalFrameML was based on its core methodology, which distinguishes between recombinant events and mutations on a phylogenetic tree, thereby accounting for the latter when quantifying the impact of recombination on EBV evolution as previously described 23 . The per-site recombination rate (⍴/bp) was then derived from the relative rate of recombination to mutation (⍴/𝛳) and average DNA length imported via recombination (∂). The gene-specific 95% highest posterior density intervals (HPDI) were derived by propagating uncertainty from the posterior distributions of ⍴/𝛳 and ∂. Genome-wide 95% HDPI for ⍴/bp were also computed for each group. Visualisation was implemented in R using ggplot, with genes ordered by their per-site recombination rate (⍴/bp) and recombination hotspots defined as genes with per-site recombination rate exceeding the upper bound of the genome-wide HPDI. Assessing EBV mutation frequency in BL compared to other phenotypes Next, we evaluated the impact of point mutations and small indels(<50bp) using the original dataset (410 genomes, including the public dataset). We assembled a comprehensive table of all variants for each phenotype. We then assigned the variants to 1-kb sliding windows based on their genomic positions. We summed the variants within each window for each sample to determine the average variant density per window and visualised this as a bar plot to identify regions with higher or lower mutation frequencies. R packages dplyr, tidyr, and ggplot were utilised for data manipulation and plotting 81 . To identify frequently mutated genes within each phenotype, we calculated the ratio of observed to expected mutations per gene. Observed mutation counts were summed per gene and phenotype, while expected counts were estimated by multiplying the genome-wide mutation rate (average variants per sample divided by genome length in kb) by each gene’s length in kb. For each gene-phenotype pair, we constructed a 2x2 contingency table, comparing mutations in the gene versus all other genes and performed Fisher’s exact tests to calculate enrichment p-values. Log2fold enrichment (log2FE) scores were computed for each gene-phenotype pair with a pseudo count of 0.5 added to avoid division by zero. P values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method 91 within each phenotype. Volcano plots were used to visualise log2fold enrichment scores and FDR-adjusted significant values. All analyses were performed in R v4.4.1 using the tidyverse, ggrepel, ggplot2 and patchwork libraries 81 . To gain a deeper understanding of EBV mutation frequency within functional gene groups, we aggregated mutation counts by gene function and compared them between BL and healthy control samples. For each gene category, group differences were assessed using the Wilcoxon rank-sum test with FDR correction for multiple testing. Cliff’s Delta 92 was calculated to estimate the effect size and direction of the difference between the BL and healthy groups. The distribution of observed mutation counts per category was visualised using violin plots with overlaid boxplots. EBV genome-wide association analysis and the risk of Burkitt lymphoma To identify genomic loci that might be associated with BL risk, we performed a genome-wide logistic regression analysis of EBV genomes from 141 BL cases and 78 healthy carriers. This approach is commonly employed in human genetic population studies to explore the genetic basis of complex diseases. Its adaptation for viral genomic research, known as viral genome-wide association studies (vGWAS), seeks to elucidate viral genetic factors that contribute to disease development or intricate interactions between the host and the virus by surveying the entire viral genome 93 . First, we merged individual sample variant caller files (VCF), split multiallelic sites introduced by VCF merging into biallelic sites using bcftools 79 . We then filtered out low-impact variants and retained high-confidence non-synonymous variants in both BL and healthy carriers. We merged this with sample metadata, which included information on sample origin, phenotype, strain, and principal components (PC1 and PC2) derived from whole-genome sequence variation to account for virus population structure. We used logistic regression with BL status (BL = 1, non-BL = 0) as the outcome and mutation presence as the binary variable indicating the presence of the alternate allele at a given locus. We adjusted the results for region, country, strain, and the first two principal components (PC1 and PC2) as potential confounding covariates. Logistic regression was performed using the glm function with the binomial family in R. Raw p-values for each variant were adjusted using the Benjamini-Hochberg false discovery rate (FDR) method to account for multiple comparisons 91 . Odds ratios (OR) were derived by exponentiating the regression coefficients, and FDR-adjusted p-values ranked variants to identify the most statistically significant associations. The Manhattan plot was used to visualise significant associations in R. De novo extraction of EBV mutation signatures and analysis To decompose the mutational signatures, we used a non-negative matrix factorisation method, a technique frequently employed in cancer genomics research 30,94 . We developed a Python pipeline to generate a 96-channel trinucleotide mutation matrix from VCF files. Single-nucleotide variants (SNVs) were annotated with their trinucleotide context from the EBV reference genome (NC_007605.1) using Biopython v1.85 87 . Mutations were normalised to the pyrimidine context, expanding the six base substitution types (C>A, C>G, C>T, T>A, T>C, T>G) into 96 trinucleotide contexts following the COSMIC convention 30 . Each sample’s VCF file was parsed with cyvcf2 v0.30.28 95 , and mutation counts were aggregated into a sample-by-context matrix, which was exported for downstream signature analysis. We extracted the EBV mutation signatures using MutSignature 94 , an R package with advanced functions for importing DNA variants, computing mutation types, and extracting mutation types via non-negative matrix factorisation (NMF) 96 . The software offers a wide range of compatibility, including the analysis of non-human genomes with support for the analysis of non-standard mutation types, such as tetranucleotide mutation types 94 . A rank survey was conducted for k = 2 to 5 across 50 iterations using both the original and randomised mutation matrices to assess signature robustness. The optimal number of signatures (k = 5) was selected based on cophenetic correlation, silhouette width, consensus clustering stability, and residual sum of squares (Suppl. Figs. 10-12). Signature exposure proportions were normalised per sample and integrated with metadata for statistical analysis. The normality of signature exposures was assessed using the Shapiro-Wilk test 97 (Suppl. Fig. 13). As the data exhibited non-normality, subsequent group comparisons employed beta regression models, with PC1, PC2, and strain as fixed effects to account for virus population structure and strain variation. Significant phenotype-SBS associations (FDR < 0.05) were reported, along with their corresponding effect estimates. We then compared the de novo EBV signatures to the well-curated COSMIC v3.2 reference SBS signatures database using cosine similarity, calculated with the lsa R package. Cosine similarity values of 0.90 or higher were considered strong matches 30 . Bar plots and heatmaps were used to visualise the best matches, following the steps in previous studies 29,30 . EBV gene expression studies To explore the relationship between EBV genomic diversity and viral gene expression, we analysed RNA from formalin-fixed paraffin-embedded (FFPE) tissue of 15 BL cases with available whole-genome sequencing data (See supplementary materials for detailed methods). Briefly, total RNA was extracted, rRNA-depleted, and used to generate Illumina-compatible libraries following standard protocols for degraded FFPE RNA. Libraries were multiplexed and sequenced on an Illumina NovaSeq XPlus platform (2 × 150 bp paired-end). Raw reads were demultiplexed with bcl2fastq. Paired-end RNA-Seq reads were first trimmed for adapters and low-quality bases using standard Illumina trimming tools. Transcript-level quantification was performed using Kallisto (v0.51.1) 98 . Reads were pseudo-aligned to the hybrid human-EBV transcriptome index to generate abundance estimates. All samples’ transcript-level abundance from Kallisto were aggregated to gene-level counts and imported using the tximport function in R together with the gene annotation. Differential gene expression (DGE) analysis was performed in DESeq2 (v1.48.1) 99 . Normalised counts were used to generate heatmaps with ComplexHeatmap (v2.24.1) 100 Declarations Parents or legal guardians of children aged 3-17 years provided written informed consent, in addition to the child's assent. Participants over 17 years old gave written informed consent before any study-related procedures were carried out. Acknowledgements We would like to thank the patients, parents, and all study participants for their invaluable contribution to this study. We are also grateful to the study teams and our collaborators across East Africa. Our special appreciation goes to Dr. Claire El Mouden, AI-REAL Programme Manager; Dr. Faraja Chiwanga, Head of Teaching and Research at Muhimbili National Hospital. We extend our gratitude to Dr. Simon Engledow at Azenta Life Sciences for preparing the FFPE RNA libraries and conducting RNA sequencing. We also thank Ms. Gagandeep Kaur Bath for coordinating the procurement and logistics necessary for the RNA sequencing. Lastly, we thank Dr. Richard Mangwi Ayiasi at Muni University for his role in coordinating the recruitment of healthy participants for the study. Author contributions AS, MDO, CC, EM, and HMM conceived the AI-REAL study. IDL designed and conducted the experiments, analysed and interpreted data and wrote the first draft of the manuscript. IDL, DV, SH and KR developed the bioinformatic pipelines, processed and analysed the whole-genome and transcriptome datasets. AB supervised laboratory work and provided technical advice on laboratory protocol development. DJ, HD, and AB designed the EBV panels for both targeted and whole-genome sequencing. HN, IO, DM, PM, EM, and PSN conducted the study recruitment and pathology specimen processing. JS, AY, EJ, and HC performed whole-genome sequencing. CA, LM, AM, and EME reviewed the pathology diagnosis of all cases recruited in the study. WFM, CC, HMM, LC, EM, and MDO were responsible for administering the study sites. KR, CSKL and AS supervised the experiments, provided technical advice on experimental design, and critically revised the manuscript. All authors read and approved the final manuscript. Competing interests AS receives honoraria from Gilead, AbbVie, Roche, Janssen, and Illumina; unrestricted educational grants from Janssen and Gilead; and in-kind contributions from Illumina and Oxford Nanopore Technologies Ltd. All other authors declare no conflict of interest. Funding This work was funded by the NIHR (NIHR-RIGHT award 200133) through UK International Development funding from the UK Government to support Global Health Research and a Ph.D. studentship from the Commonwealth Scholarship Commission. The views expressed in this publication are those of the authors and do not necessarily reflect the views of the sponsors or funders. Data availability Extended Data Fig.1 Extended Data Fig.2 Extended Data Fig.3 Supplementary Data 1 Information on analysis samples Supplementary Data 2 Log2fold mutation enrichment in all samples Raw sequencing data (FASTQs) and novel EBV genomes will be uploaded to the appropriate biorepositories. Additional information Further information on study design, methods (including detailed testing protocols), statistical tables, and figures is available in the supplementary information. References Vetsika, E.-K. & Callan, M. Infectious mononucleosis and Epstein-Barr virus. Expert reviews in molecular medicine 6 , 1-16 (2004). Dunmire, S.K., Verghese, P.S. & Balfour, H.H. Primary Epstein-Barr virus infection. Journal of Clinical Virology 102 , 84-92 (2018). Münz, C. Latency and lytic replication in Epstein–Barr virus-associated oncogenesis. Nature Reviews Microbiology 17 , 691-700 (2019). Wong, Y., Meehan, M.T., Burrows, S.R., Doolan, D.L. & Miles, J.J. Estimating the global burden of Epstein-Barr virus-related cancers. J Cancer Res Clin Oncol 148 , 31-46 (2022). Bjornevik, K., Münz, C., Cohen, J.I. & Ascherio, A. Epstein-Barr virus as a leading cause of multiple sclerosis: mechanisms and implications. Nat Rev Neurol 19 , 160-171 (2023). Houen, G. & Trier, N.H. Epstein-Barr Virus and Systemic Autoimmune Diseases. Frontiers in Immunology 11 (2021). Hämmerl, L., Colombet, M., Rochford, R., Ogwang, D.M. & Parkin, D.M. The burden of Burkitt lymphoma in Africa. Infect Agent Cancer 14 , 17 (2019). Yu, H. et al. The global burden of nasopharyngeal carcinoma from 2009 to 2019: an observational study based on the Global Burden of Disease Study 2019. European Archives of Oto-Rhino-Laryngology 279 , 1519-1533 (2022). Kirimunda, S. et al. Variation in the Human Leukocyte Antigen system and risk for endemic Burkitt lymphoma in northern Uganda. British Journal of Haematology 189 , 489-499 (2020). Kanda, T., Yajima, M. & Ikuta, K. Epstein‐Barr virus strain variation and cancer. Cancer science 110 , 1132-1139 (2019). Burkitt, D.P. Etiology of Burkitt's lymphoma—an alternative hypothesis to a vectored virus. Journal of the National Cancer Institute 42 , 19-28 (1969). Bahri, R. et al. Epstein-Barr Virus (EBV) is mostly latent and clonal in angioimmunoblastic T cell lymphoma (AITL). Cancers 14 , 2899 (2022). Pathmanathan, R., Prasad, U., Sadler, R., Flynn, K. & Raab-Traub, N. Clonal proliferations of cells infected with Epstein–Barr virus in preinvasive lesions related to nasopharyngeal carcinoma. New England Journal of Medicine 333 , 693-698 (1995). Young, L.S., Arrand, J.R. & Murray, P.G. EBV gene expression and regulation. in Human Herpesviruses: Biology, Therapy, and Immunoprophylaxis (eds. Arvin, A. et al. ) (Cambridge University Press, Cambridge, 2007). Murata, T. et al. Molecular basis of epstein–barr virus latency establishment and lytic reactivation. Viruses 13 , 2344 (2021). Inagaki, T. et al. Direct evidence of abortive lytic infection-mediated establishment of Epstein-Barr virus latency during B-cell infection. Frontiers in Microbiology 11 , 575255 (2021). Xue, S.a. et al. Promiscuous expression of Epstein‐Barr virus genes in Burkitt's lymphoma from the central African country Malawi. International journal of cancer 99 , 635-643 (2002). Abate, F. et al. Distinct Viral and Mutational Spectrum of Endemic Burkitt Lymphoma. PLOS Pathogens 11 , e1005158 (2015). Agwati, E.O. et al. Profiling genome-wide recombination in Epstein Barr virus reveals type-specific patterns and associations with endemic-Burkitt lymphoma. Virology Journal 19 , 208 (2022). Yao, Q. et al. Isolation of intertypic recombinants of Epstein-Barr virus from T-cell-immunocompromised individuals. Journal of virology 70 , 4895-4903 (1996). Pérez-Losada, M., Arenas, M., Galán, J.C., Palero, F. & González-Candelas, F. Recombination in viruses: mechanisms, methods of study, and evolutionary consequences. Infect Genet Evol 30 , 296-307 (2015). Berenstein, A.J., Lorenzetti, M.A. & Preciado, M.V. Recombination rates along the entire Epstein Barr virus genome display a highly heterogeneous landscape. Infection, Genetics and Evolution 65 , 96-103 (2018). Zanella, L. et al. A reliable Epstein-Barr Virus classification based on phylogenomic and population analyses. Scientific Reports 9 , 9829 (2019). Xu, M. et al. Genome sequencing analysis identifies Epstein–Barr virus subtypes associated with high risk of nasopharyngeal carcinoma. Nature Genetics 51 , 1131-1136 (2019). Rickinson, A.B., Young, L.S. & Rowe, M. Influence of the Epstein-Barr virus nuclear antigen EBNA 2 on the growth phenotype of virus-transformed B cells. J Virol 61 , 1310-7 (1987). Sample, J. et al. Epstein-Barr virus types 1 and 2 differ in their EBNA-3A, EBNA-3B, and EBNA-3C genes. Journal of Virology 64 , 4084-4092 (1990). Hui, K.F. et al. High risk Epstein‐Barr virus variants characterized by distinct polymorphisms in the EBER locus are strongly associated with nasopharyngeal carcinoma. International journal of cancer 144 , 3031-3042 (2019). da Costa, V.G., Marques-Silva, A.C. & Moreli, M.L. The Epstein-Barr virus latent membrane protein-1 (LMP1) 30-bp deletion and XhoI-polymorphism in nasopharyngeal carcinoma: a meta-analysis of observational studies. Systematic Reviews 4 , 46 (2015). Alexandrov, L.B., Nik-Zainal, S., Wedge, D.C., Campbell, P.J. & Stratton, M.R. Deciphering signatures of mutational processes operative in human cancer. Cell reports 3 , 246-259 (2013). Alexandrov, L.B. et al. The repertoire of mutational signatures in human cancer. Nature 578 , 94-101 (2020). Godsmark, G., Rios, L.A.D.S. & Mowla, S. Activation-induced cytidine deaminase promotes proliferation and enhances chemoresistance and migration in B-cell lymphoma. Anticancer Research 41 , 237-247 (2021). Shilova, O., Tsyba, D. & Shilov, E. Mutagenic activity of AID/APOBEC deaminases in antiviral defense and carcinogenesis. Molecular Biology 56 , 46-58 (2022). Wan, J.C. et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nature Reviews Cancer 17 , 223-238 (2017). Xu, M. et al. Genome-wide profiling of Epstein-Barr virus integration by targeted sequencing in Epstein-Barr virus associated malignancies. Theranostics 9 , 1115 (2019). Xiao, K. et al. Genome-wide analysis of Epstein-Barr virus (EBV) integration and strain in C666-1 and Raji cells. Journal of Cancer 7 , 214 (2016). Martin, D.P., Murrell, B., Golden, M., Khoosal, A. & Muhire, B. RDP4: Detection and analysis of recombination patterns in virus genomes. Virus Evolution 1 (2015). Sausen, D., Poirier, M., Spiers, L. & Smith, E. Mechanisms of T cell evasion by Epstein-Barr virus and implications for tumor survival. Frontiers in immunology 14 , 1289313 (2023). Ressing, M.E. et al. Epstein-Barr virus evasion of CD8+ and CD4+ T cell immunity via concerted actions of multiple gene products. in Seminars in cancer biology Vol. 18 397-408 (Elsevier, 2008). Zuo, J. et al. The Epstein-Barr virus G-protein-coupled receptor contributes to immune evasion by targeting MHC class I molecules for degradation. PLoS pathogens 5 , e1000255 (2009). Jochum, S., Moosmann, A., Lang, S., Hammerschmidt, W. & Zeidler, R. The EBV immunoevasins vIL-10 and BNLF2a protect newly infected B cells from immune recognition and elimination. PLoS pathogens 8 , e1002704 (2012). Yap, L.F., Wong, A.K.C., Paterson, I.C. & Young, L.S. Functional implications of Epstein-Barr virus lytic genes in carcinogenesis. Cancers 14 , 5780 (2022). Young, L.S. & Rickinson, A.B. Epstein–Barr virus: 40 years on. Nature Reviews Cancer 4 , 757-768 (2004). Balfour Jr, H.H. et al. Age-specific prevalence of Epstein–Barr virus infection among individuals aged 6–19 years in the United States and factors affecting its acquisition. The Journal of infectious diseases 208 , 1286-1293 (2013). De-Thé, G. et al. Epidemiological evidence for causal relationship between Epstein-Barr virus and Burkitt's lymphoma from Ugandan prospective study. Nature 274 , 756-761 (1978). Briercheck, E.L. et al. Geographic EBV variants confound disease-specific variant interpretation and predict variable immune therapy responses. Blood Advances 8 , 3731-3744 (2024). Chiara, M. et al. Geographic population structure in Epstein-Barr virus revealed by comparative genomics. Genome biology and evolution 8 , 3284-3291 (2016). Palser, A.L. et al. Genome diversity of Epstein-Barr virus from multiple tumor types and normal infection. Journal of virology 89 , 5222-5237 (2015). Patiño-Galindo, J.Á., Filip, I. & Rabadan, R. Global patterns of recombination across human viruses. Molecular biology and evolution 38 , 2520-2531 (2021). Levitskaya, J., Sharipo, A., Leonchiks, A., Ciechanover, A. & Masucci, M.G. Inhibition of ubiquitin/proteasome-dependent protein degradation by the Gly-Ala repeat domain of the Epstein-Barr virus nuclear antigen 1. Proc Natl Acad Sci U S A 94 , 12616-21 (1997). Xue, S.A. & Griffin, B.E. Complexities associated with expression of Epstein-Barr virus (EBV) lytic origins of DNA replication. Nucleic Acids Res 35 , 3391-406 (2007). Mancao, C. & Hammerschmidt, W. Epstein-Barr virus latent membrane protein 2A is a B-cell receptor mimic and essential for B-cell survival. Blood 110 , 3715-21 (2007). Chen, X., Kamranvar, S.A. & Masucci, M.G. Oxidative stress enables Epstein-Barr virus-induced B-cell transformation by posttranscriptional regulation of viral and cellular growth-promoting factors. Oncogene 35 , 3807-16 (2016). Kikuchi, K. et al. Overexpression of Activation-Induced Cytidine Deaminase in MTX- and Age-Related Epstein-Barr Virus-Associated B-Cell Lymphoproliferative Disorders of the Head and Neck. J Oncol 2015 , 605750 (2015). Lirussi, L. & Nilsen, H.L. DNA glycosylases define the outcome of endogenous base modifications. International Journal of Molecular Sciences 24 , 10307 (2023). Gunnarsson, R. et al. Single base substitution and insertion/deletion mutational signatures in adult core binding factor acute myeloid leukemia. Leukemia 36 , 1681-1684 (2022). Yaacov, A. et al. Cancer mutational signatures identification in clinical assays using neural embedding-based representations. Cell Rep Med 5 , 101608 (2024). Karihtala, P., Kilpivaara, O. & Porvari, K. Mutational signatures and their association with survival and gene expression in urological carcinomas. Neoplasia 44 , 100933 (2023). Ponomarev, G.V. et al. APOBEC mutagenesis is low in most types of non-B DNA structures. iScience 25 , 104535 (2022). Tomkova, M., Tomek, J., Kriaucionis, S. & Schuster-Böckler, B. Mutational signature distribution varies with DNA replication timing and strand asymmetry. Genome Biology 19 , 129 (2018). Vasimuddin, M., Misra, S., Li, H. & Aluru, S. Efficient architecture-aware acceleration of BWA-MEM for multicore systems. in 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS) 314-324 (IEEE, 2019). Koboldt, D.C. et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome research 22 , 568-576 (2012). Lai, Z. et al. VarDict: a novel and versatile variant caller for next-generation sequencing in cancer research. Nucleic acids research 44 , e108-e108 (2016). Benjamin, D. et al. Calling somatic SNVs and indels with Mutect2. Biorxiv , 861054 (2019). Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. fly 6 , 80-92 (2012). Prjibelski, A., Antipov, D., Meleshko, D., Lapidus, A. & Korobeynikov, A. Using SPAdes De Novo Assembler. Current Protocols in Bioinformatics 70 , e102 (2020). Gurevich, A., Saveliev, V., Vyahhi, N. & Tesler, G. QUAST: quality assessment tool for genome assemblies. Bioinformatics 29 , 1072-1075 (2013). Camacho, C. et al. BLAST+: architecture and applications. BMC bioinformatics 10 , 421 (2009). Zhou, W. et al. ABACUS: An Electronic Structure Analysis Package for the AI Era. arXiv preprint arXiv:2501.08697 (2025). Li, H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34 , 3094-3100 (2018). Capella-Gutiérrez, S., Silla-Martínez, J.M. & Gabaldón, T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25 , 1972-1973 (2009). Katoh, K., Misawa, K., Kuma, K.i. & Miyata, T. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic acids research 30 , 3059-3066 (2002). Smit, A. Smit, AFA, Hubley, R & Green, P. RepeatMasker Open-3.0. (2015). Togkousidis, A., Kozlov, O.M., Haag, J., Höhler, D. & Stamatakis, A. Adaptive RAxML-NG: accelerating phylogenetic inference under maximum likelihood using dataset difficulty. Molecular Biology and Evolution 40 , msad227 (2023). Lemoine, F. et al. Renewing Felsenstein’s phylogenetic bootstrap in the era of big data. Nature 556 , 452-456 (2018). Paradis, E. & Schliep, K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics 35 , 526-528 (2019). Yu, G., Smith, D.K., Zhu, H., Guan, Y. & Lam, T.T.Y. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods in ecology and evolution 8 , 28-36 (2017). Xu, S. et al. ggtreeExtra: compact visualization of richly annotated phylogenetic data. Molecular biology and evolution 38 , 4039-4042 (2021). Wickham, H. et al. Welcome to the Tidyverse. Journal of open source software 4 , 1686 (2019). Danecek, P. et al. Twelve years of SAMtools and BCFtools. GigaScience 10 (2021). Chang, C.C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 4 , s13742-015-0047-8 (2015). Team, R.C. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www . R-project. org/ (2016). Romero, E.V. & Feder, A.F. Elevated HIV viral load is associated with higher recombination rate in vivo. Molecular Biology and Evolution 41 , msad260 (2024). Vos, P.G. et al. Evaluation of LD decay and various LD-decay estimators in simulated and SNP-array data of tetraploid potato. Theor Appl Genet 130 , 123-135 (2017). Castresana, J. Selection of Conserved Blocks from Multiple Alignments for Their Use in Phylogenetic Analysis. Molecular Biology and Evolution 17 , 540-552 (2000). Robinson, D.F. & Foulds, L.R. Comparison of phylogenetic trees. Mathematical biosciences 53 , 131-147 (1981). Schliep, K.P. phangorn: phylogenetic analysis in R. Bioinformatics 27 , 592-593 (2011). Cock, P.J.A. et al. Biopython: freely available Python tools for computational molecular biology and bioinformatics. Bioinformatics 25 , 1422-1423 (2009). Team, T.P.D. pandas-dev/pandas: Pandas. Zenodo, February (2020). Wong, T.K. et al. IQ-TREE 3: Phylogenomic Inference Software using Complex Evolutionary Models. (2025). Didelot, X. & Wilson, D.J. ClonalFrameML: efficient inference of recombination in whole bacterial genomes. PLoS computational biology 11 , e1004041 (2015). Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological) 57 , 289-300 (1995). Meissel, K. & Yao, E. Using Cliff’s Delta as a Non-Parametric Effect Size Measure: An Accessible Web App and R Tutorial. 2024. (1977). Butković, A. & Elena, S.F. Genome-wide association studies of viral infections—A short guide to a successful experimental and statistical analysis. Frontiers in Systems Biology 2 , 1005758 (2022). Fantini, D., Vidimar, V., Yu, Y., Condello, S. & Meeks, J.J. MutSignatures: an R package for extraction and analysis of cancer mutational signatures. Scientific Reports 10 , 18217 (2020). Pedersen, B.S. & Quinlan, A.R. Who’s who? Detecting and resolving sample anomalies in human DNA sequencing studies with peddy. The American Journal of Human Genetics 100 , 406-413 (2017). Gaujoux, R. & Seoighe, C. A flexible R package for nonnegative matrix factorization. BMC bioinformatics 11 , 367 (2010). Shapiro, S.S. & Wilk, M.B. An analysis of variance test for normality (complete samples). Biometrika 52 , 591-611 (1965). Bray, N.L., Pimentel, H., Melsted, P. & Pachter, L. Near-optimal probabilistic RNA-seq quantification. Nature biotechnology 34 , 525-527 (2016). Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology 15 , 550 (2014). Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32 , 2847-2849 (2016). Additional Declarations Yes there is potential Competing Interest. None Supplementary Files SupplementaryInformation.pdf Supplementary Information ExtendedDataFig1.png ExtendedDataFig2.png ExtendedDataFig3.png Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-7473738\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Article\",\"associatedPublications\":[],\"authors\":[{\"id\":511723191,\"identity\":\"5de74321-d275-44d6-90db-d0926e142a0d\",\"order_by\":0,\"name\":\"Ismail Legason\",\"email\":\"data:image/png;base64,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\",\"orcid\":\"https://orcid.org/0000-0002-6118-542X\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Ismail\",\"middleName\":\"\",\"lastName\":\"Legason\",\"suffix\":\"\"},{\"id\":511723192,\"identity\":\"72287da5-6d28-4b6f-ad02-280f17dda250\",\"order_by\":1,\"name\":\"Adam Burns\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-8787-9147\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Adam\",\"middleName\":\"\",\"lastName\":\"Burns\",\"suffix\":\"\"},{\"id\":511723193,\"identity\":\"9ead2277-38e0-4f33-bf2f-89f9d50ce9df\",\"order_by\":2,\"name\":\"Dimitrios Vavoulis\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-3984-1507\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Dimitrios\",\"middleName\":\"\",\"lastName\":\"Vavoulis\",\"suffix\":\"\"},{\"id\":511723194,\"identity\":\"1bb77390-faf0-44a4-9695-d40ea0c111e5\",\"order_by\":3,\"name\":\"Silvia Halim\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Silvia\",\"middleName\":\"\",\"lastName\":\"Halim\",\"suffix\":\"\"},{\"id\":511723195,\"identity\":\"df89f788-e515-482d-bef6-9301b5682cba\",\"order_by\":4,\"name\":\"Helene Dréau\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Helene\",\"middleName\":\"\",\"lastName\":\"Dréau\",\"suffix\":\"\"},{\"id\":511723196,\"identity\":\"b3f103bb-a1fd-4c61-989a-c3d091245cec\",\"order_by\":5,\"name\":\"Daisy Jennings\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Daisy\",\"middleName\":\"\",\"lastName\":\"Jennings\",\"suffix\":\"\"},{\"id\":511723197,\"identity\":\"e9e85c5a-9723-4e1d-ba6b-3a99a5ee7ad6\",\"order_by\":6,\"name\":\"Hadijah Nabalende\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"African Field Epidemiology Network\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Hadijah\",\"middleName\":\"\",\"lastName\":\"Nabalende\",\"suffix\":\"\"},{\"id\":511723198,\"identity\":\"32cf691e-964b-46ff-89bf-b866c047c19e\",\"order_by\":7,\"name\":\"Isaac Otim\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"EMBLEM Study, St. Mary’s Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Isaac\",\"middleName\":\"\",\"lastName\":\"Otim\",\"suffix\":\"\"},{\"id\":511723199,\"identity\":\"525bbab7-5d98-4886-89b6-9c851ddb2627\",\"order_by\":8,\"name\":\"Martin Ogwang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"St. Mary's Hospital, Lacor\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Martin\",\"middleName\":\"\",\"lastName\":\"Ogwang\",\"suffix\":\"\"},{\"id\":511723200,\"identity\":\"e109c49a-8ad7-4bb9-b0d3-0d16a6b4686d\",\"order_by\":9,\"name\":\"Julius Sseruyange\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Central Public Health Laboratory(CPHL)\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Julius\",\"middleName\":\"\",\"lastName\":\"Sseruyange\",\"suffix\":\"\"},{\"id\":511723201,\"identity\":\"0ccce8c5-1a47-4b29-991d-6010a6b08dff\",\"order_by\":10,\"name\":\"Alisen Ayitewala\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Central Public Health Laboratory(CPHL)\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Alisen\",\"middleName\":\"\",\"lastName\":\"Ayitewala\",\"suffix\":\"\"},{\"id\":511723202,\"identity\":\"f313a925-cbee-4685-9d1a-ec1098d38568\",\"order_by\":11,\"name\":\"Caroline Achola\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Central Public Health Laboratory(CPHL)\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Caroline\",\"middleName\":\"\",\"lastName\":\"Achola\",\"suffix\":\"\"},{\"id\":511723203,\"identity\":\"c808d2b8-fe69-4fbe-828c-1a95f3cf2738\",\"order_by\":12,\"name\":\"Susan Nabadda\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Central Public Health Laboratory(CPHL)\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Susan\",\"middleName\":\"\",\"lastName\":\"Nabadda\",\"suffix\":\"\"},{\"id\":511723204,\"identity\":\"c6eeb5c6-6c24-4c82-9f51-5999178096a1\",\"order_by\":13,\"name\":\"Emmanuel Josephat\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili University of Health and Allied Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Emmanuel\",\"middleName\":\"\",\"lastName\":\"Josephat\",\"suffix\":\"\"},{\"id\":511723205,\"identity\":\"217873b1-deb6-4a11-bd0d-a6f92fe477f5\",\"order_by\":14,\"name\":\"Heavenlight Christopher\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili University of Health and Allied Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Heavenlight\",\"middleName\":\"\",\"lastName\":\"Christopher\",\"suffix\":\"\"},{\"id\":511723206,\"identity\":\"1d94ada7-ad8e-49d1-8589-d20dbfc6b9b1\",\"order_by\":15,\"name\":\"William Mawalla\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili University of Health and Allied Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"William\",\"middleName\":\"\",\"lastName\":\"Mawalla\",\"suffix\":\"\"},{\"id\":511723207,\"identity\":\"a72ce471-ec18-42ba-8957-df90bb9d652e\",\"order_by\":16,\"name\":\"Clara Chamba\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili University of Health and Allied Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Clara\",\"middleName\":\"\",\"lastName\":\"Chamba\",\"suffix\":\"\"},{\"id\":511723208,\"identity\":\"55448092-f331-49a6-bee5-8091d4866ff6\",\"order_by\":17,\"name\":\"Eric Magorosa\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili National Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Eric\",\"middleName\":\"\",\"lastName\":\"Magorosa\",\"suffix\":\"\"},{\"id\":511723209,\"identity\":\"2ac3b9c1-6a63-4c86-9514-ce155a50b3fe\",\"order_by\":18,\"name\":\"Leah Mnango\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili National Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Leah\",\"middleName\":\"\",\"lastName\":\"Mnango\",\"suffix\":\"\"},{\"id\":511723210,\"identity\":\"acf19c4f-149a-4f7b-9a9c-dd1ccad4351b\",\"order_by\":19,\"name\":\"Lulu Chirande\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili National Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Lulu\",\"middleName\":\"\",\"lastName\":\"Chirande\",\"suffix\":\"\"},{\"id\":511723211,\"identity\":\"a15cbb59-c521-44d0-aec1-26b72b22b0e1\",\"order_by\":20,\"name\":\"Hadija mwamtemi\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Muhimbili National Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Hadija\",\"middleName\":\"\",\"lastName\":\"mwamtemi\",\"suffix\":\"\"},{\"id\":511723212,\"identity\":\"d6ff78e3-0a49-4f35-8971-e98eb6577bdd\",\"order_by\":21,\"name\":\"Daniel Mbwambo\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Kilimanjaro Christian Medical Centre\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Daniel\",\"middleName\":\"\",\"lastName\":\"Mbwambo\",\"suffix\":\"\"},{\"id\":511723213,\"identity\":\"b82dee6f-3aca-4ad5-a4bb-00c8feb89ff9\",\"order_by\":22,\"name\":\"Priscus Mapendo\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Kilimanjaro Christian Medical Centre\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Priscus\",\"middleName\":\"\",\"lastName\":\"Mapendo\",\"suffix\":\"\"},{\"id\":511723214,\"identity\":\"40d97a8d-061d-4b50-a1a5-47d18d0ed9f0\",\"order_by\":23,\"name\":\"Alex Mremi\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0001-7226-0168\",\"institution\":\"Kilimanjaro Christian Medical Centre\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Alex\",\"middleName\":\"\",\"lastName\":\"Mremi\",\"suffix\":\"\"},{\"id\":511723215,\"identity\":\"ed23f83c-b80e-4961-9e68-3e261b16fec4\",\"order_by\":24,\"name\":\"Elifuraha Mkwizu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Kilimanjaro Christian Medical Centre University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Elifuraha\",\"middleName\":\"\",\"lastName\":\"Mkwizu\",\"suffix\":\"\"},{\"id\":511723216,\"identity\":\"7694f1d5-f7f8-4bee-82c3-71d1bc62e54d\",\"order_by\":25,\"name\":\"Paul Ntemi\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Bugando Medical Centre\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Paul\",\"middleName\":\"\",\"lastName\":\"Ntemi\",\"suffix\":\"\"},{\"id\":511723217,\"identity\":\"bc635259-693a-4b29-ab43-ff31137cc217\",\"order_by\":26,\"name\":\"Heronima Joas\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Bugando Medical Centre\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Heronima\",\"middleName\":\"\",\"lastName\":\"Joas\",\"suffix\":\"\"},{\"id\":511723218,\"identity\":\"342052ad-ffdb-421f-b743-ced4329ab323\",\"order_by\":27,\"name\":\"Edrick Elias\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Catholic University of Health and Allied Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Edrick\",\"middleName\":\"\",\"lastName\":\"Elias\",\"suffix\":\"\"},{\"id\":511723219,\"identity\":\"5fe7b6b8-bc35-4c59-868c-44a189ed397e\",\"order_by\":28,\"name\":\"Carol Leung\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-1545-7228\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Carol\",\"middleName\":\"\",\"lastName\":\"Leung\",\"suffix\":\"\"},{\"id\":511723220,\"identity\":\"388d0e5c-a361-4374-923c-2b2b32632682\",\"order_by\":29,\"name\":\"Kate Ridout\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Oxford\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Kate\",\"middleName\":\"\",\"lastName\":\"Ridout\",\"suffix\":\"\"},{\"id\":511723221,\"identity\":\"91c3399b-266f-438e-b0b0-0f4fcb4ec9c2\",\"order_by\":30,\"name\":\"Anna Schuh\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-3938-8490\",\"institution\":\"Oxford University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Anna\",\"middleName\":\"\",\"lastName\":\"Schuh\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-08-27 17:01:13\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-7473738/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-7473738/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":93777774,\"identity\":\"88f971cb-e006-4ac4-98c6-17cb50692c6b\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":660686,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003ePhylogenetic analysis of 410 EBV genomes\\u003c/p\\u003e\\n\\u003cp\\u003eThe heatmaps: the outer ring represents sample phenotypes, the middle ring denotes the country of origin, and the inner ring indicates the region. The phenotypes analysed in this study include BL, EBV-positive healthy individuals, HL, NKTCL, NPC, and other rare EBV-associated malignancies. The tips are annotated with EBV type (type 1 green triangle; type 2 blue dot), and the most recent ancestor (tree root) is indicated. The support values shown at the nodes represent TBE values computed from RAxML-ng; nodes with support values greater than 70% are displayed.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/3a0dc147074f2712a9a0b869.png\"},{\"id\":93777775,\"identity\":\"adb55f9f-af4a-42de-98ae-70c7bdff9b67\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":210833,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003ePrincipal Component Analysis (PCA) of EBV genomes (n =410)\\u003c/p\\u003e\\n\\u003cp\\u003eThe scatter plot illustrates the genetic variation in EBV genomes. The first two principal components account for 18.8% and 15.4% of the variation, respectively. Points are coloured by region, with corresponding phenotype annotations shown as a heatmap below the scatter plot. Additional tracks display normalised PC1 and PC2 values (z-scores), emphasising systematic differences between regions and related disease phenotypes. Samples from Africa (red) cluster closely on the right side (high PC1). Samples from Asia (green) cluster on the lower left (low PC1, low PC2). American samples (Central/North/South) exhibit greater dispersion, primarily to the left on PC1, yet remain high on PC2, thereby distinguishing Asian strains from American strains.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/2d7eff2fafd13160e2c5b579.png\"},{\"id\":93777783,\"identity\":\"25a3c0aa-989c-4d22-942c-ac1b8432c27f\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":184290,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eComparative analysis of linkage disequilibrium (LD) decay across sample groups\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eA) \\u003c/strong\\u003eBoxplot showing the distribution of mean pairwise LD (measured as r\\u003csup\\u003e2\\u003c/sup\\u003e) per sample, stratified by group. Each box represents the interquartile range (IQR), with the median shown as a horizontal black line. \\u003cstrong\\u003eB) \\u003c/strong\\u003eBootstrapped LOESS curves showing LD decay with 95% confidence intervals per group. Shaded ribbons represent uncertainty from individual-level resampling, illustrating variability in decay rates. Sample groups included Burkitt lymphoma (BL), EBV-positive healthy controls, nasopharyngeal carcinoma (NPC), Hodgkin lymphoma (HL), natural killer T-cell lymphoma (NKTCL), and other EBV-associated malignancies.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/9089ba2b9f39f8d6addf48c8.png\"},{\"id\":93777777,\"identity\":\"8ccf086f-4c6c-4bb7-a9e8-08111a25c083\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":342918,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eGene-wise recombination rates (ρ/bp) in EBV coding genes from Burkitt Lymphoma (BL) and EBV-positive healthy controls\\u003c/p\\u003e\\n\\u003cp\\u003eEach point represents the posterior mean of ρ per base pair (ρ/bp) for a gene, as estimated by ClonalFrameML. Top panel: EBV genes from BL. Bottom panel: EBV genes from EBV-positive healthy controls. Genes are ordered by their recombination rate. Vertical lines indicate the 95% highest posterior density intervals (HPDIs). The shaded grey area in each panel represents the genome-wide 95% HPDI for ρ/bp. Genes are colour-coded across panels, with each gene assigned a unique colour to highlight concordance or divergence in recombination across host groups. Statistical significance is considered when the gene’s HPDI interval does not overlap with the genome-wide 95% HPDI (grey area). The gene-specific error bars show intragenic heterogeneity in the recombination rate. The large 95% HPDI reflects more variability in the ρ/bp signal.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/05ce933850f561bcde0af488.png\"},{\"id\":93777780,\"identity\":\"287c3f41-ed32-4474-b45a-3687ecc78725\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":250692,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eMean single-nucleotide variant (SNV) density across the EBV genome in Burkitt lymphoma\\u003c/p\\u003e\\n\\u003cp\\u003eThe line plot shows the mean SNV count per 1 kb window across the EBV genome in BL. Red dots indicate hypervariable regions (significantly elevated SNV density), while blue dots represent SNV deserts (conserved regions). Genes corresponding to hypervariable and conserved regions are labelled. The gene track was annotated using the B95.8 (NC_007605.1) reference genome. Directional arrows represent EBV genes, coloured by functional group (Latency, Early, Immediate-Early, Late, Immune Evasion, Uncharacterized).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/a4a27239f44883e6e1b61899.png\"},{\"id\":93777784,\"identity\":\"e8c52718-1b74-47ac-8320-71469de8468f\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":6,\"title\":\"Figure 6\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":216937,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eBeta regression model of EBV mutational signature proportions in disease phenotypes relative to healthy controls\\u003c/p\\u003e\\n\\u003cp\\u003eThe plot displays effect size estimates (β coefficients) from beta regression models assessing the association between EBV mutational signatures and phenotypes. Bars show the direction and magnitude of effects relative to the reference group (healthy controls), with positive values indicating increased signature activity and negative values indicating decreased activity. Effect sizes (β coefficients) with 95% confidence intervals are shown. Phenotypes analysed included Burkitt lymphoma (BL), EBV-positive healthy controls, Hodgkin lymphoma (HL), NKTCL, nasopharyngeal carcinoma (NPC) and other EBV-associated conditions.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/634c256fb330e7f461ebe6dc.png\"},{\"id\":93779070,\"identity\":\"9573eafe-0b59-44db-a238-2208d99f0dd4\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:50:57\",\"extension\":\"png\",\"order_by\":7,\"title\":\"Figure 7\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":256857,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eThe top five COSMIC SBS signature matches to de novo EBV signatures\\u003c/p\\u003e\\n\\u003cp\\u003eThe bar plots show cosine similarity between each EBV-derived mutational signature (SBS_EBV1 to SBS_EBV5) and the top five matches to COSMIC single-base substitution (SBS) signatures. The x-axis represents the cosine similarities, and the y-axis represents the Cosmic SBS signatures. The highest similarity was observed for SBS_EBV1, SBS_EBV4, and SBS_EBV5, which closely matched COSMIC SBS30, SBS1, and SBS45 with a cosine similarity greater than 0.9. While SBS_EBV2 and SBS_EBV3 showed weak matches to COSMIC SBS 42 and 54, suggesting distinct mutational signatures in EBV-associated malignancies.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/97593bd2050fb286c5066773.png\"},{\"id\":93777785,\"identity\":\"3bec71da-b0d5-4986-9e4e-f8acb8f6673d\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:42:57\",\"extension\":\"png\",\"order_by\":8,\"title\":\"Figure 8\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":306130,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eExpression levels of mutated EBV genes in Burkitt lymphoma cases\\u003c/p\\u003e\\n\\u003cp\\u003eThe plot illustrates the expression levels of EBV genes identified as mutated in the study. Normalised counts of the EBV genes are shown for 15 BL tumours analysed through FFPE ribosome-depleted total RNA sequencing. The grey violin and boxplot display the expression levels of 10 selected human housekeeping genes (ACTB, GAPDH, HPRT1, B2M, PP1A, RPLP0, RPL13A, TBP, GUSB, and SDHA). The red dashed line marks a normalised count of 1. Latent and immune evasion genes are coloured, while lytic genes remain uncoloured to aid comparison.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/be6bdf3d7900fe0d6947bdd2.png\"},{\"id\":93781294,\"identity\":\"78c8e4fd-c31d-48b8-bc70-0719cd65a967\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 13:14:59\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":3766403,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/b5a92dcb-fdd1-4c9a-ac41-964199e15b76.pdf\"},{\"id\":93780102,\"identity\":\"67089361-45ef-482f-817b-3facfa2013ba\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:58:57\",\"extension\":\"pdf\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":4852348,\"visible\":true,\"origin\":\"\",\"legend\":\"Supplementary Information\",\"description\":\"\",\"filename\":\"SupplementaryInformation.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/9f39019dcd0d4459012c7fdd.pdf\"},{\"id\":93779073,\"identity\":\"766ad790-a0df-4011-92c9-08d42be708ae\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:50:57\",\"extension\":\"png\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":3084616,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"ExtendedDataFig1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/eae1a298d8965d2127a7c2e2.png\"},{\"id\":93780457,\"identity\":\"ab5b6c89-9bcf-4a89-9a8a-0f59affd504e\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 13:06:57\",\"extension\":\"png\",\"order_by\":3,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":1059613,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"ExtendedDataFig2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/da826d1703dbf76772d61d93.png\"},{\"id\":93779071,\"identity\":\"23e860e8-52f5-4398-864b-f619bac8c374\",\"added_by\":\"auto\",\"created_at\":\"2025-10-17 12:50:57\",\"extension\":\"png\",\"order_by\":4,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":210500,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"ExtendedDataFig3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7473738/v1/4cceb908721491e47cbbeb15.png\"}],\"financialInterests\":\"\\u003cb\\u003eYes\\u003c/b\\u003e there is potential Competing Interest.\\nNone\",\"formattedTitle\":\"Evidence of latent-lytic replication in EBV-positive Burkitt lymphoma from whole genome and transcriptome sequencing\",\"fulltext\":[{\"header\":\"Introduction\",\"content\":\"\\u003cp\\u003eThe Epstein-Barr virus (EBV), a double-stranded DNA virus of the γ-herpesvirus family, is the most ubiquitous human pathogen, best known for causing infectious mononucleosis\\u003csup\\u003e\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u003c/sup\\u003e. Human infections occur through saliva, with at least 90% of the global population exposed to the virus at some point during their lifetime\\u003csup\\u003e\\u003cspan additionalcitationids=\\\"CR2\\\" citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e\\u003c/sup\\u003e. While most infections are asymptomatic, approximately 1% of cases may lead to complications ranging from neoplastic disorders to systemic autoimmune diseases \\u003csup\\u003e\\u003cspan additionalcitationids=\\\"CR4 CR5\\\" citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e\\u003c/sup\\u003e.\\u003c/p\\u003e\\u003cp\\u003eAmong EBV-associated neoplasms, Burkitt lymphoma (BL) and nasopharyngeal carcinoma (NPC) represent the two well-characterised models \\u003csup\\u003e\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e\\u003c/sup\\u003e. BL, an aggressive B-cell malignancy, is primarily a paediatric malignancy and highly endemic in equatorial Africa and parts of Oceania, with an estimated annual incidence of 0.17 to 6.2 per 100,000 children in some parts of equatorial Africa \\u003csup\\u003e\\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e\\u003c/sup\\u003e. In contrast, NPC, an epithelial malignancy, predominantly affects adults and is highly endemic in Southeast Asia and parts of North Africa, with an incidence of 0.3 to 2.1 cases per 100,000 individuals \\u003csup\\u003e\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e\\u003c/sup\\u003e. Although both malignancies share a strong link to EBV, their distinct geographic distributions, patient age, environmental cofactors (\\u003cem\\u003eP. falciparum\\u003c/em\\u003e for BL\\u003csup\\u003e\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e\\u003c/sup\\u003e, and dietary or lifestyle factors for NPC \\u003csup\\u003e\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e\\u003c/sup\\u003e), as well as previously described host-specific genetic factors\\u003csup\\u003e\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e\\u003c/sup\\u003e, point towards a complex interplay of all these variables in determining these divergent disease phenotypes. Notably, the role of specific EBV genome characteristics remains largely elusive.\\u003c/p\\u003e\\u003cp\\u003eThe EBV genome is generally regarded as stable, evolving clonally in most tumours\\u003csup\\u003e\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e\\u003c/sup\\u003e. After B-cell infection, EBV persists latently during clonal proliferation, germinal centre transit and memory B-cell differentiation, with \\u003cem\\u003eEBNA-1\\u003c/em\\u003e as the primary antigen tethering episomal genomes to the host chromatin\\u003csup\\u003e\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e\\u003c/sup\\u003e. Occasional abortive reactivation events, producing a restricted set of lytic proteins without full viral production, have been proposed as a mechanism for long-term persistence\\u003csup\\u003e\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e16\\u003c/span\\u003e\\u003c/sup\\u003e. However, emerging evidence indicates that a full lytic reactivation of the viral genome may be more frequent in BL than previously recognised \\u003csup\\u003e\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e\\u003c/sup\\u003e, a process that can amplify viral mutations and promote recombination, thereby accelerating viral evolution.\\u003c/p\\u003e\\u003cp\\u003eRecombination appears to be more common in settings of high EBV co-infection rates, such as endemic regions\\u003csup\\u003e\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e\\u003c/sup\\u003e, and in immunodeficient individuals\\u003csup\\u003e\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e\\u003c/sup\\u003e, generating mosaic genomes through homologous or non-homologous exchange of genetic material between co-infecting strains\\u003csup\\u003e\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e\\u003c/sup\\u003e. While recombination rates vary across herpesviruses\\u003csup\\u003e\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e\\u003c/sup\\u003e, studies specifically focused on BL remain limited\\u003csup\\u003e\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e\\u003c/sup\\u003e, and it is not yet clear whether recombination or mutation is the dominant force driving EBV diversification in this context. Mutational processes, including point mutations, small deletions, and polymorphisms, introduce substantial genetic variation with profound biological consequences\\u003csup\\u003e\\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e\\u003c/sup\\u003e. For example, \\u003cem\\u003eEBNA-2\\u003c/em\\u003e sequence variation defines viral subtypes and modulates B-cell transformation\\u003csup\\u003e\\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e\\u003c/sup\\u003e, while variants of \\u003cem\\u003eLMP-1, BALF2\\u003c/em\\u003e, and \\u003cem\\u003eRPMS1\\u003c/em\\u003e are linked to poor clinical outcomes in NPC-endemic regions\\u003csup\\u003e\\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e27\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e28\\u003c/span\\u003e\\u003c/sup\\u003e. Furthermore, intertypic recombinants are frequently identified in certain EBV-associated conditions\\u003csup\\u003e\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e\\u003c/sup\\u003e. Collectively, these observations suggest that both recombination and mutation act as complementary forces shaping EBV genomic diversity, underscoring the need for comparative analyses of their relative contributions to BL pathogenesis.\\u003c/p\\u003e\\u003cp\\u003eWhile recombination events can produce large genomic mosaics\\u003csup\\u003e\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e\\u003c/sup\\u003e, mutations frequently show distinctive signatures that reveal their underlying mechanisms\\u003csup\\u003e\\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e\\u003c/sup\\u003e. As such, it is vital to understand both the processes driving EBV evolution and the mutational mechanisms behind them, as this knowledge can help identify biomarkers that may be useful for diagnosis or prognostic assessment. In cancer genomics, mutational signature analysis has proven invaluable for identifying footprints of distinct mutagenic processes. For example, smoking, UV-light or enzymatic activities such as AID- and APOBEC-mediated cytidine deamination, oxidative stress, or defective mismatch all cause specific mutational footprints\\u003csup\\u003e\\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e30\\u003c/span\\u003e\\u003c/sup\\u003e. EBV-infected B cells undergo somatic hypermutations in germinal centres, where AID is naturally expressed and APOBEC activity may also be elevated \\u003csup\\u003e\\u003cspan citationid=\\\"CR31\\\" class=\\\"CitationRef\\\"\\u003e31\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e\\u003c/sup\\u003e, offering a plausible endogenous mechanism of viral mutation. The APOBEC family comprises cytidine deaminases that primarily edit cellular mRNA, but also act on replicating viruses, forming part of the host innate immune response \\u003csup\\u003e\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e\\u003c/sup\\u003e. Applying a mutational signature framework to EBV genomes may therefore reveal whether host-derived pressures, such as germinal centre-associated somatic hypermutation or APOBEC activity, leave a discernible imprint on the viral genome. This approach can provide deeper insight into the interplay between host immunity, viral evolution, and oncogenesis.\\u003c/p\\u003e\\u003cp\\u003eTo explore these hypotheses, we analysed 64 novel EBV genomes isolated from cell-free DNA (cfDNA) samples, along with 346 geographically representative public genomes, all of which were isolates from primary clinical samples. Our cohort included children and young adults with and without BL in Uganda and Tanzania. cfDNA comprises apoptotic and necrotic cell products released into the bloodstream\\u003csup\\u003e\\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e33\\u003c/span\\u003e\\u003c/sup\\u003e. In neoplastic disease, these products may include tumour fragments, hence the term \\u0026lsquo;circulating tumour DNA\\u0026rsquo;. Because the EBV genome is frequently integrated into the host genome \\u003csup\\u003e\\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e34\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e35\\u003c/span\\u003e\\u003c/sup\\u003e, analysing circulating tumour DNA offers a non-invasive window into the tumour-associated viral genome.\\u003c/p\\u003e\\u003cp\\u003eThe present analysis aimed to understand the relative contribution of recombination and mutation events to EBV genome diversification in BL compared to other EBV-associated conditions by 1) describing differences in the viral genome across geographic regions and disease phenotypes, 2) assessing the recombination frequency of EBV genomes derived from BL compared to other phenotypes, 3) identifying viral variants enriched in BL, 4) characterising mutational signatures associated with BL using the COSMIC single-base substitution (SBS) model, and 5) linking viral genetic alterations, including recombination and mutational patterns to EBV gene expression programmes in BL.\\u003c/p\\u003e\"},{\"header\":\"Results\",\"content\":\"\\u003cp\\u003eStudy cohort characteristics\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eAn overview of the study is provided in the Supplementary Methods (Suppl. Figs. 1-3). Briefly, participants were identified from the Aggressive Infection-Related East African Lymphoma (AI-REAL) study, a prospective case-control study in Uganda and Tanzania. Out of a 310 sample-study cohort, 53, including BL and other lymphoma samples for whom sufficient EBV ctDNA was available, underwent whole genome sequencing. Eleven EBV-positive healthy controls with sufficient EBV DNA levels were included in the final sample cohort, resulting in a total of 64 primary study samples. \\u0026nbsp;The median age of the study cohort was 9 years. There were more males than females (64% vs. 36%), and 77% of the study cohort came from Uganda. Clinical symptoms such as weight loss, cachexia, and night sweats were common among cases, with over half of the individuals diagnosed with advanced-stage tumours (Suppl. Table 1).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eEBV whole genome sequencing\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eWe sequenced the 64 EBV whole genomes using a capture-based protocol. The average sequencing depth across the runs was 271.7x in the present study, with 92.6% of the EBV genome covered by at least 10x coverage (Suppl. Fig. 4). A total of 36,963 variants (SNVs and indels \\u0026lt; 50 bp) were identified across the 64 genomes. Per sample average, there were 641 variants in BL compared to 625 in other malignancies and 290 in EBV-positive healthy controls. To verify the accuracy of sequencing and variant calling, we processed samples in duplicate and called variants to assess the reproducibility. The results showed that the coefficient of variation (CV) between the first and second run was less than 4% (Suppl. Table 2).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eSummary of the published genomes\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eNext, we combined data from these new samples with published data to create a dataset of 410 EBV whole genomes, comprising 332 samples from individuals with malignancies and 78 from EBV-positive healthy controls, all derived from EBV DNA extracted either from primary tumour samples or plasma (for information on analysis samples, see Supplementary Data 1). Most of the public EBV genomes lacked strain information. Using the Basic Local Alignment Search Tool (BLAST) approach, we were able to identify 266 (76.9%) of these genomes as Type 1 and 80 (23.1%) as Type 2 (Suppl. Fig. 5). Mixed or undetermined strains were excluded from further analysis. This analysis was vital in ensuring we ordered the contigs obtained from the de novo assemblies against the correct EBV reference genome. \\u0026nbsp;The mean sequencing depth for the public data set ranged from 141.8x to 776.92x, and genome coverage varied between 75.0% and 98.6% across studies. We prioritised samples with at least 10x depth and 75% genome coverage for our downstream analysis (Suppl. Fig. 4). \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eEBV genomes display population structure shaped by region, with no evident phenotypic clusters\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eWe analysed EBV genome diversity using phylogenetic and dimensionality reduction techniques. A total of 410 EBV genomes, including reference genomes (NC_007605.1 for EBV Type 1 and NC_009334.1 for Type 2) and two well-characterised EBV-positive B cell lines, Daudi and Raji, were used to build the phylogenetic tree. These genomes originate from five regions: Africa (Kenya, Malawi, Uganda), Asia (China), Central America (Guatemala, Peru), North America, and South America (Suppl. Fig. 3) \\u0026nbsp;\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe phylogenetic results revealed a clear geographic population structure of the EBV genomes, clustering by region regardless of the disease phenotypes involved (Fig.1). Within African samples, we observed no distinct patterns between cases and EBV-positive healthy controls. Similar patterns were also observed in Asian samples, with no evident phenotypic clustering. Out of 141 BL samples, two were from Asia, and these formed a subcluster with Asian NPCs; the African BL cases clustered with healthy African samples. In terms of strains, the larger proportion of African type 2 EBV genomes formed a distinct subcluster from type 1 genomes, irrespective of the disease phenotype. Interestingly, four of the NPC-associated EBV genomes from Asia clustered with type 2 African EBV genomes. Two of these were previously reported as Type 1, and the other two were not assigned any genotype in the original study\\u003csup\\u003e24\\u003c/sup\\u003e. \\u0026nbsp;However, when we performed alignment (BLAST) against \\u003cem\\u003eEBNA2\\u003c/em\\u003e-type-specific contigs, these genomes showed the strongest hits for type 2 EBV. \\u0026nbsp;Hence, we categorised them into type 2 genomes.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe heatmaps: the outer ring represents sample phenotypes, the middle ring denotes the country of origin, and the inner ring indicates the region. The phenotypes analysed in this study include BL, EBV-positive healthy individuals, HL, NKTCL, NPC, and other rare EBV-associated malignancies. The tips are annotated with EBV type (type 1 green triangle; type 2 blue dot), and the most recent ancestor (tree root) is indicated. The support values shown at the nodes represent TBE values computed from RAxML-ng; nodes with support values greater than 70% are displayed.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTo corroborate our phylogenetic results, we conducted principal component analysis on the 410 genomes. A total of 36,100 variants were loaded from the binary PLINK file. Thirty-four thousand three hundred twenty-nine (34,329) variants were excluded due to minor allele thresholds (VAF \\u0026lt; 0.05), and an additional 995 were pruned due to linkage disequilibrium (r\\u0026sup2; \\u0026gt; 0.6), resulting in a final set of 776 variants used to generate the relationship matrix. In the PCA, the origin of samples or genomes contributed most to the genetic variation (Fig. 2). \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe first two principal components (PC1 \\u0026amp; PC2), which explained ~ 34% of the total variance, showed a strong clustering by geography (Africa vs. Asia vs. the Americas). PC1 separated African strains from Asian strains, accounting for 18.8% of the variation. PC2 explained 15.4% of the genetic variation and distinguished American EBV strains from those of Asian and African origins. Consistent with the phylogeny, no apparent phenotypic clustering was observed. EBV strains (Type 1 vs Type 2) were dispersed across regions and did not appear to define principal axes of variation. Even in higher-order PCs (PC3-PC6) (Suppl. Fig. 6), the clustering was weak by strain, suggesting that intra-strain variation is minor compared to regional diversity.\\u003c/p\\u003e\\n\\u003cp\\u003eThe scatter plot illustrates the genetic variation in EBV genomes. The first two principal components account for 18.8% and 15.4% of the variation, respectively. Points are coloured by region, with corresponding phenotype annotations shown as a heatmap below the scatter plot. Additional tracks display normalised PC1 and PC2 values (z-scores), emphasising systematic differences between regions and related disease phenotypes. Samples from Africa (red) cluster closely on the right side (high PC1). Samples from Asia (green) cluster on the lower left (low PC1, low PC2). American samples (Central/North/South) exhibit greater dispersion, primarily to the left on PC1, yet remain high on PC2, thereby distinguishing Asian strains from American strains.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eIncreased EBV recombination in BL compared to EBV-positive healthy controls and other EBV-associated malignancies, as suggested by linkage disequilibrium analysis\\u003c/p\\u003e\\n\\u003cp\\u003eTo explore differences in the degree of recombination events contributing to the clonal evolution of EBV in disease phenotypes, we calculated pairwise linkage disequilibrium (LD) correlation coefficients (r\\u0026sup2;) between genomic sites within a 10 kb sliding window for up to 99,999 single-nucleotide variants. Our results revealed steeper LD decay patterns for BL (Fig. 3), NPC, and EBV-positive healthy controls, with mean r\\u0026sup2; ranging from 0.3 to 0.45 over genomic distances of ~500 bp to 5000 bp, compared to HL, NKTCL, and other EBV-associated malignancies, which maintained higher mean r\\u0026sup2; values over longer distances before decaying. At longer distances, e.g., \\u0026gt; 5000 bp, the mean r\\u0026sup2; values generally plateaued (mean r\\u0026sup2; ~ 0.3). At the longest distances (~10,000 bp), most groups converged to low mean r\\u0026sup2; values. \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eImpact of recombination on EBV genome diversification in BL compared to EBV-positive healthy controls\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eGiven the strong recombination signals observed in BL-derived EBV genomes, we investigated whether these events were uniformly distributed across the viral genome or concentrated in specific regions, and how they contributed to viral diversification compared with EBV-positive healthy carriers. We analysed 141 BL cases and 78 EBV-positive healthy controls. Whole-genome alignments (trimmed for poorly aligned regions and gaps) were screened using RDP4\\u003csup\\u003e36\\u003c/sup\\u003e, identifying putative recombination in most isolates, with 43 events reaching statistical significance (Suppl. Table 3). To assess their impact, we reconstructed phylogenies of the 219 EBV genomes with and without recombinant regions. The topologies differed markedly, with a normalised Robinson\\u0026ndash;Foulds distance of 0.884, confirming strong recombination-driven incongruence (Extended Data Fig. 1).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eWe next quantified the relative contribution of recombination to EBV genome diversification\\u0026ndash;accounting for mutations using ClonalFrameML\\u003csup\\u003e23\\u003c/sup\\u003e. Gene-level estimates showed substantial heterogeneity, with wide 95% highest posterior density intervals (HPDIs; Fig. 4). Two latency genes, \\u003cem\\u003eLMP-1\\u003c/em\\u003e and \\u003cem\\u003eEBNA-1\\u003c/em\\u003e, displayed recombination rates well above the genome average in BL, whereas the tegument gene \\u003cem\\u003eBOLF1\\u003c/em\\u003e was enriched for recombination in healthy carriers. Other genes had rates near background levels with overlapping HPDIs and no statistical significance.\\u003c/p\\u003e\\n\\u003cp\\u003eEach point represents the posterior mean of \\u0026rho; per base pair (\\u0026rho;/bp) for a gene, as estimated by ClonalFrameML. Top panel: EBV genes from BL. Bottom panel: EBV genes from EBV-positive healthy controls. Genes are ordered by their recombination rate. Vertical lines indicate the 95% highest posterior density intervals (HPDIs). The shaded grey area in each panel represents the genome-wide 95% HPDI for \\u0026rho;/bp. Genes are colour-coded across panels, with each gene assigned a unique colour to highlight concordance or divergence in recombination across host groups. Statistical significance is considered when the gene\\u0026rsquo;s HPDI interval does not overlap with the genome-wide 95% HPDI (grey area). The gene-specific error bars show intragenic heterogeneity in the recombination rate. The large 95% HPDI reflects more variability in the \\u0026rho;/bp signal. \\u0026nbsp;\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eEBV latency and immune evasion genes are significantly mutated in BL compared to other EBV-associated phenotypes\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo assess the impact of mutations on EBV genome diversity and the risk of endemic BL, we quantified the frequency of single-nucleotide variants (SNVs) and indels across the viral genome. A total of 410 EBV genomes from BL, HL, NPC, NKTCL, EBV-positive healthy controls, and other EBV-associated malignancies were analysed. Our results showed that the per-sample mutation frequency of the EBV genome was similar in BL and EBV-positive healthy controls (4.3 per kb), but higher in NKTCL (7.98 per kb), NPC (5.3 per kb), HL (4.8 per kb), and other cancers (4.4 per kb), respectively (Suppl. Fig.7). However, genome-wide analysis of EBV single nucleotide variants \\u0026nbsp;(SNV) in BL samples (141 cases) \\u0026nbsp;revealed significant regions of sequence variation (hypervariable) and conservation (Fig. 5). The hypervariable regions corresponded to key immune evasion genes such as \\u003cem\\u003eLMP1\\u003c/em\\u003e\\u0026ndash;oncogenic signalling\\u003csup\\u003e37\\u003c/sup\\u003e,\\u003cem\\u003e\\u0026nbsp;EBNA-1\\u003c/em\\u003e, \\u0026nbsp;\\u003cem\\u003eBNLF2a\\u003c/em\\u003e (TAP inhibitor\\u003csup\\u003e38\\u003c/sup\\u003e), \\u003cem\\u003eBILF2 (\\u003c/em\\u003eviral G protein-coupled receptor \\u003csup\\u003e39\\u003c/sup\\u003e), and \\u003cem\\u003eBCRF1\\u003c/em\\u003e (viral interleukin-10\\u003csup\\u003e40\\u003c/sup\\u003e), as well as lytic genes \\u0026nbsp;\\u003cem\\u003eBBLF4\\u003c/em\\u003e (component of viral helicase primase\\u003cem\\u003e), BLRF1\\u003c/em\\u003e (gN), \\u003cem\\u003eBBRF3\\u003c/em\\u003e (gM), \\u003cem\\u003eBGLF3\\u003c/em\\u003e and \\u003cem\\u003eRPMS1\\u003c/em\\u003e (BART transcripts)\\u003csup\\u003e41\\u003c/sup\\u003e. The conserved regions corresponded to the \\u003cem\\u003eEBNA-1\\u003c/em\\u003e Gly-Ala, \\u003cem\\u003eLF3\\u003c/em\\u003e \\u003cem\\u003e(IR4)\\u003c/em\\u003e, and the LMP2A terminal repeats \\u003cem\\u003e(TR)\\u003c/em\\u003e of the viral genome.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe line plot shows the mean SNV count per 1 kb window across the EBV genome in BL. Red dots indicate hypervariable regions (significantly elevated SNV density), while blue dots represent SNV deserts (conserved regions). Genes corresponding to hypervariable and conserved regions are labelled. The gene track was annotated using the B95.8 (NC_007605.1) reference genome. Directional arrows represent EBV genes, coloured by functional group (Latency, Early, Immediate-Early, Late, Immune Evasion, Uncharacterized). \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eWe next performed gene-level mutation enrichment analysis across six EBV-associated phenotypes (BL, NPC, HL, NKTCL, other cancers, and EBV-positive healthy controls). Mutation fold enrichment was calculated as log₂(observed/expected) per gene (Supplementary Data 2). Latency genes showed the strongest enrichment in lymphoma phenotypes (except HL), with BL exhibiting the broadest and highest fold changes, particularly in EBNA-1, EBNA-LP, EBNA-3A, LMP-1, and LMP-2A/2B (e.g., LMP-1: log₂FE = 7.5 in BL vs. 6.7 in NPC and 5.9 in healthy controls; Extended Data Fig. 2). In contrast, structural genes (e.g., BLLF1, BLLF2, BDLF3, BPLF1, BRRF2, BNRF1, BGLF5, BGLF1) showed limited or sporadic enrichment. When grouped by functional categories (capsid, envelope, immune evasion, latency, replication, tegument), pairwise comparisons between BL and healthy controls revealed no significant differences (Suppl. Fig. 8). Cliff\\u0026rsquo;s delta indicated small-to-moderate effect sizes, with the strongest difference in immune-evasion genes (\\u0026Delta; = 0.556).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTo complement this gene-level analysis, we performed a genome-wide association study (GWAS) of 933 polymorphic loci in 141 BL cases and 78 EBV-positive healthy controls. Logistic regression adjusted for viral population structure (PC1, PC2), geography, and strain revealed no variants reaching genome-wide (p \\u0026lt; 5 \\u0026times; 10⁻⁸) or suggestive (p \\u0026lt; 1 \\u0026times; 10⁻⁵) significance. Several loci showed nominal associations (p \\u0026lt; 0.05) but did not warrant fine mapping (Suppl. Fig. 9, Table 4).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eEnrichment of G-rich mutation signatures in EBV genomes from endemic Burkitt lymphoma\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eHaving established that EBV genomes from BL samples exhibited a higher mutational burden compared to EBV-positive healthy controls, we next sought to investigate the mechanisms underlying this diversification. To this end, we applied non-negative matrix factorisation (NMF) to deconvolute the mutational processes shaping the EBV genome diversification. This analysis identified five distinct mutational signatures (SBS_EBV1\\u0026ndash;SBS_EBV5) across BL, EBV-positive healthy controls, Hodgkin lymphoma (HL), NK/T-cell lymphoma (NKTCL), nasopharyngeal carcinoma (NPC), and other EBV-associated malignancies (Extended Data Fig. 3). SBS_EBV1 showed strong enrichment of C \\u0026gt; T transitions, particularly at C-rich trinucleotide contexts, e.g., C[C\\u0026gt;T]C, T[C\\u0026gt;T]C, C[C\\u0026gt;T]A and C[C\\u0026gt;T]T. SBS_EBV2 showed dominance in C\\u0026gt;T and T\\u0026gt;C transitions with a bias towards G- and C-rich flanks, e.g., G[C\\u0026gt;T]C, C[T\\u0026gt;C]G, G[T\\u0026gt;C]C, G[C\\u0026gt;T]G). SBS_EBV3 was dominated by T\\u0026gt;C and C\\u0026gt;T transitions in G-rich contexts, with peak contributions at C[T\\u0026gt;C]G, G[T\\u0026gt;C]G, and A[T\\u0026gt;C]G. SBS_EBV4 showed a strong bias for C\\u0026gt;T transitions at CpG sites, while SBS_EBV5 \\u0026nbsp;was enriched in C\\u0026gt;A transversions at CpC, CpA, and CpG contexts. Phenotype-specific enrichment analysis revealed notable differences (Fig. 6): SBS_EBV1 was significantly enriched in NKTCL; SBS_EBV2 was largely absent from disease phenotypes but detectable in EBV-positive healthy carriers; SBS_EBV3 was significantly enriched in BL relative to all other groups; SBS_EBV4 showed moderate activity in HL and other rare malignancies but was near background in BL, NPC, and NKTCL; and SBS_EBV5 was broadly active across malignancies. \\u0026nbsp;\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTo relate these EBV signatures (SBS_EBV1\\u0026ndash;5) \\u0026nbsp;to known mutational processes in human cancers, we calculated the cosine similarities to the COSMIC SBS v3.2 reference set (Fig. 7). SBS_EBV1 showed the highest similarity with SBS30 (cosine similarity = 0.90), SBS_EBV4 and SBS_EBV5 were highly similar to SBS1 and SBS45 (cosine similarity \\u0026gt; 0.9), respectively. By contrast, SBS_EBV2 and SBS_EBV3 demonstrated only modest similarity to known signatures, with SBS42 (cosine similarity = 0.65) and SBS54 (cosine similarity = 0.74) as their closest matches. SBS_EBV3 constituted a distinct mutational spectrum in BL samples, suggesting a BL-specific mutagenic process. No EBV signatures matched the canonical APOBEC-associated COSMIC signatures SBS2 and SBS13, which are characterised by predominance of C \\u0026gt; T and C \\u0026gt; G substitutions in TpCpW trinucleotide contexts\\u003csup\\u003e30\\u003c/sup\\u003e.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe plot displays effect size estimates (\\u0026beta; coefficients) from beta regression models assessing the association between EBV mutational signatures and phenotypes. Bars show the direction and magnitude of effects relative to the reference group (healthy controls), with positive values indicating increased signature activity and negative values indicating decreased activity. Effect sizes (\\u0026beta; coefficients) with 95% confidence intervals are shown. Phenotypes analysed included Burkitt lymphoma (BL), EBV-positive healthy controls, Hodgkin lymphoma (HL), NKTCL, nasopharyngeal carcinoma (NPC) and other EBV-associated conditions.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe bar plots show cosine similarity between each EBV-derived mutational signature (SBS_EBV1 to SBS_EBV5) and the top five matches to COSMIC single-base substitution (SBS) signatures. The x-axis represents the cosine similarities, and the y-axis represents the Cosmic SBS signatures. \\u0026nbsp;The highest similarity was observed for SBS_EBV1, SBS_EBV4, and SBS_EBV5, which closely matched COSMIC SBS30, SBS1, and SBS45 with a cosine similarity greater than 0.9. While SBS_EBV2 and SBS_EBV3 showed weak matches to COSMIC SBS 42 and 54, suggesting distinct mutational signatures in EBV-associated malignancies.\\u003c/p\\u003e\\n\\u003cp\\u003eHeterogeneous latent-lytic reactivation and replication events of EBV in endemic Burkitt lymphoma \\u0026nbsp;\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eBuilding\\u0026nbsp;upon our observations of an increased recombination rate, significant numbers of mutations in latency and immune evasion genes,\\u003cstrong\\u003e\\u0026nbsp;\\u003c/strong\\u003eand the enrichment of a distinct mutational signature (SBS_EBV3) in BL-associated EBV genomes, we sought to understand how these reflected on the viral transcriptional landscape. The use of primary tumour tissues from 15 additional patients for this analysis also allowed us to independently assess our previous findings from circulating tumour DNA sequencing. \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eIn addition to the noncoding EBV transcripts (\\u003cem\\u003eEBER-1\\u003c/em\\u003e and \\u003cem\\u003eEBER-2\\u003c/em\\u003e), EBV genomes in BL displayed relatively higher expression levels of both latent and lytic genes in most tumour samples analysed (Suppl. Fig. 14). Thirteen out of fifteen tumour samples (86.7%) showed detectable expression of two or more latency-associated genes. \\u003cem\\u003eEBNA-1\\u003c/em\\u003e, characteristic of the latency I programme in BL, was detected in 9/15 (60%) of the samples. Latency II-associated genes \\u003cem\\u003eLMP-1\\u003c/em\\u003e, \\u003cem\\u003eLMP-2A\\u003c/em\\u003e or \\u003cem\\u003eB\\u003c/em\\u003e were expressed in 12/15 (80%) of tumours, while latency III-associated genes \\u003cem\\u003eEBNA-2\\u003c/em\\u003e, \\u0026nbsp;\\u003cem\\u003eEBNA-3A\\u003c/em\\u003e, \\u003cem\\u003eEBNA-3B\\u003c/em\\u003e and \\u003cem\\u003eEBNA-3C\\u003c/em\\u003e were detectable in over 60% (9/15) of the samples. Notably, 80% (12/15) of the tumours exhibited detectable expression of lytic genes alongside latent genes, suggesting that full lytic replication occurs in most of the BL tumours analysed. The lytic switch gene \\u003cem\\u003eBZLF1\\u003c/em\\u003e was detectable in 10 out of 15 (66.7%) of the samples. \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eNotably, most of the latent and immune evasion EBV genes (\\u003cem\\u003eLMP-1, EBNA-1, EBNA-2, EBNA-3A, EBNA-3B, EBNA-3C, EBNA-LP, BCRF1, BILF2, and BZLF1\\u003c/em\\u003e) and lytic genes that we identified as significantly affected by mutations were actively transcribed in BL tumours (Fig. 8)\\u003cem\\u003e.\\u0026nbsp;\\u003c/em\\u003eOverall, these results suggest a non-canonical EBV latency programme in BL, with most viral episomes initiating lytic reactivation, as indicated by \\u003cem\\u003eBZLF1\\u003c/em\\u003e expression in the majority of tumour samples analysed. \\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe plot illustrates the expression levels of EBV genes identified as mutated in the study. Normalised counts of the EBV genes are shown for 15 BL tumours analysed through FFPE ribosome-depleted total RNA sequencing. The grey violin and boxplot display the expression levels of 10 selected human housekeeping genes (ACTB, GAPDH, HPRT1, B2M, PP1A, RPLP0, RPL13A, TBP, GUSB, and SDHA). The red dashed line marks a normalised count of 1. Latent and immune evasion genes are coloured, while lytic genes remain uncoloured to aid comparison. \\u0026nbsp;\\u003c/p\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eThe Epstein-Barr virus (EBV) is one of the most widespread human pathogens, typically persisting as a lifelong latent infection\\u003csup\\u003e\\u003cspan additionalcitationids=\\\"CR43\\\" citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e42\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR44\\\" class=\\\"CitationRef\\\"\\u003e44\\u003c/span\\u003e\\u003c/sup\\u003e. Although EBV genomes in tumours were once regarded as clonally evolving\\u003csup\\u003e\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e\\u003c/sup\\u003e, our analysis of 410 genomes, including 64 newly sequenced genomes from BL patients and EBV-positive healthy individuals in East Africa, reveals a more complex and dynamic evolutionary landscape. Consistent with previous findings\\u003csup\\u003e\\u003cspan additionalcitationids=\\\"CR46\\\" citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e45\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR47\\\" class=\\\"CitationRef\\\"\\u003e47\\u003c/span\\u003e\\u003c/sup\\u003e, the EBV population was shaped primarily by geography rather than disease phenotype, suggesting founder effects, host-driven selection, and recurrent population bottlenecks. Within the African cohorts, both type 1 and type 2 EBV clusters were observed, though regional signatures dominated over viral subtype, underscoring the influence of host-virus co-evolution.\\u003c/p\\u003e\\u003cp\\u003eRecombination emerged as a dominant force in EBV evolution, consistent with previous research\\u003csup\\u003e\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e\\u003c/sup\\u003e. Linkage disequilibrium (LD) analysis revealed elevated recombination rates in BL compared to healthy carriers and other EBV-associated malignancies. Conversely, LD decay was slower in NKTCL and HL, suggesting reduced recombination\\u003csup\\u003e\\u003cspan citationid=\\\"CR48\\\" class=\\\"CitationRef\\\"\\u003e48\\u003c/span\\u003e\\u003c/sup\\u003e. Phylogenetic incongruence and genome-wide signals further supported recombination as a pervasive driver of EBV evolution, particularly enriched in immune evasion and latency genes such as \\u003cem\\u003eEBNA-1\\u003c/em\\u003e and \\u003cem\\u003eLMP-1\\u003c/em\\u003e in BL cases, highlighting the role of host immune pressure and tumour microenvironment in selecting recombinants with enhanced fitness\\u003csup\\u003e\\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e37\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e\\u003c/sup\\u003e. In parallel, EBV genomes displayed a heterogeneous mutational landscape, with hotspots in latency-associated and immune evasion genes (e.g., \\u003cem\\u003eLMP-1\\u003c/em\\u003e\\u003csup\\u003e\\u003cem\\u003e37\\u003c/em\\u003e\\u003c/sup\\u003e, \\u003cem\\u003eBNLFa\\u003c/em\\u003e\\u003csup\\u003e\\u003cem\\u003e\\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e\\u003c/em\\u003e\\u003c/sup\\u003e, \\u003cem\\u003eBILF2\\u003c/em\\u003e\\u003csup\\u003e\\u003cem\\u003e39\\u003c/em\\u003e\\u003c/sup\\u003e, and \\u003cem\\u003eBCRF1\\u003c/em\\u003e\\u003csup\\u003e\\u003cem\\u003e40\\u003c/em\\u003e\\u003c/sup\\u003e\\u003cem\\u003e)\\u003c/em\\u003e, but conservation in regions essential for viral genome maintenance (\\u003cem\\u003eEBNA-1\\u003c/em\\u003e Gly-Ala repeat\\u003csup\\u003e\\u003cspan citationid=\\\"CR49\\\" class=\\\"CitationRef\\\"\\u003e49\\u003c/span\\u003e\\u003c/sup\\u003e, \\u003cem\\u003eLF3\\u003c/em\\u003e-IR\\u003csup\\u003e\\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e50\\u003c/span\\u003e\\u003c/sup\\u003e, and \\u003cem\\u003eLMP-2A\\u003c/em\\u003e\\u003csup\\u003e\\u003cem\\u003e\\u003cspan citationid=\\\"CR51\\\" class=\\\"CitationRef\\\"\\u003e51\\u003c/span\\u003e\\u003c/em\\u003e\\u003c/sup\\u003e). The frequent recombination rates and mutational burden in immune-modulatory genes suggest an active role for viral diversification in tumour immune escape \\u003csup\\u003e\\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e\\u003c/sup\\u003e.\\u003c/p\\u003e\\u003cp\\u003eMutational signature analysis identified a novel BL-enriched profile (SBS_EBV3), distinct from known COSMIC single-base substitution (SBS) signatures. SBS_EBV3, characterised by T\\u0026thinsp;\\u0026gt;\\u0026thinsp;C and C\\u0026thinsp;\\u0026gt;\\u0026thinsp;T substitutions in G-rich sequence contexts, may reflect oxidative DNA damage, potentially amplified by EBV-induced reactive oxygen species (ROS) via \\u003cem\\u003eLMP-1\\u003c/em\\u003e activity\\u003csup\\u003e\\u003cspan citationid=\\\"CR52\\\" class=\\\"CitationRef\\\"\\u003e52\\u003c/span\\u003e\\u003c/sup\\u003e and cytidine deamination mediated by AID overexpression\\u003csup\\u003e\\u003cspan citationid=\\\"CR53\\\" class=\\\"CitationRef\\\"\\u003e53\\u003c/span\\u003e\\u003c/sup\\u003e. This signature may highlight a previously unrecognised EBV-specific mutational process contributing to BL pathogenesis. In contrast, SBS-EBV2 was absent from tumours but present in healthy carriers, raising the possibility of a protective effect. Other EBV signatures resembled known COSMIC SBS signatures: SBS_EBV1 for SBS30 (base excision repair defects)\\u003csup\\u003e\\u003cspan citationid=\\\"CR54\\\" class=\\\"CitationRef\\\"\\u003e54\\u003c/span\\u003e\\u003c/sup\\u003e, SBS_EBV4 for SBS1 (spontaneous 5-methylcytosine deamination)\\u003csup\\u003e\\u003cspan citationid=\\\"CR55\\\" class=\\\"CitationRef\\\"\\u003e55\\u003c/span\\u003e\\u003c/sup\\u003e, and SBS_EBV5 for SBS45\\u0026ndash;aetiology uncertain but linked to sequencing artefacts in one study\\u003csup\\u003e\\u003cspan citationid=\\\"CR56\\\" class=\\\"CitationRef\\\"\\u003e56\\u003c/span\\u003e\\u003c/sup\\u003e and another associated it with better outcomes in urological carcinomas\\u003csup\\u003e\\u003cspan citationid=\\\"CR57\\\" class=\\\"CitationRef\\\"\\u003e57\\u003c/span\\u003e\\u003c/sup\\u003e. Importantly, no APOBEC-associated signatures (SBS2 and SBS13)\\u003csup\\u003e\\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e30\\u003c/span\\u003e\\u003c/sup\\u003e were detected, suggesting that APOBEC editing is not a dominant force in EBV evolution in these cohorts; however, subtle contributions cannot be excluded without motif and strand-bias-aware analyses\\u003csup\\u003e\\u003cspan citationid=\\\"CR58\\\" class=\\\"CitationRef\\\"\\u003e58\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR59\\\" class=\\\"CitationRef\\\"\\u003e59\\u003c/span\\u003e\\u003c/sup\\u003e.\\u003c/p\\u003e\\u003cp\\u003eTranscriptional profiling showed that BL-associated EBV genomes did not strictly follow the traditional latency I programme long associated with the disease \\u003csup\\u003e\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e\\u003c/sup\\u003e. Instead, they exhibited diverse expression patterns of latent and lytic genes, suggesting mixed or \\u0026lsquo;intermediate\\u0026rsquo; transcriptional states. This transcriptional plasticity may facilitate recombination and the accumulation of mutations, aligning with the enrichment of mutations in immune evasion genes and the emergence of SBS_EBV3. This dynamic viral activity challenges the classical binary model of EBV latency in BL. Instead, it supports the idea that latency exists along a functional continuum influenced by complex viral-host interactions in tumour-specific contexts.\\u003c/p\\u003e\\u003cp\\u003eThis study has several limitations that should be considered when interpreting our EBV genome and transcriptome findings. First, although we prioritised high-quality FASTQ data, sequencing artefacts and platform-specific differences could have introduced spurious mutational signatures. We mitigated this through stringent filtering and by requiring consensus across two variant callers. Second, the absence of host genotype data prevented us from controlling for host genetic influences on viral evolution or disease susceptibility, limiting our ability to fully disentangle viral from host\\u0026ndash;viral interaction effects. Third, the lack of patient-level covariates (e.g., age, sex) in public datasets further limited our ability to adjust for potential confounders. Fourth, the cross-sectional design using archival samples precludes causal inference; we cannot determine whether observed viral differences precede disease or result from tumour evolution and selection pressures. Geographic variation in EBV diversity was partly addressed by including multiple regions and adjusting for viral population structure, but the absence of paired host ancestry data limited our ability to fully correct for population stratification. Finally, transcriptome profiling from FFPE material is inherently constrained by RNA degradation and chemical modification, which may bias coverage and detection sensitivity. We minimised these effects by applying RNA integrity/DV200 thresholds for input selection, using FFPE-optimised library preparation, and applying rigorous downstream normalisation, quality filtering and batch correction.\\u003c/p\\u003e\\u003cp\\u003eDespite these limitations, this study provides the most comprehensive analysis to date of EBV genomic variation in BL. By integrating phylogenetic, mutational, and transcriptional data, we show that EBV in BL evolves through a combination of recombination, mutation, and transcriptional heterogeneity, yielding distinct genomic signatures not observed in other EBV-associated malignancies. These findings expand our understanding of EBV biology and open new avenues for diagnostics and interventions, including vaccines targeting both latent and lytic viral antigens in EBV-associated Burkitt lymphoma. Future studies integrating host-virus data with functional validation will be essential to establish the diagnostic and prognostic significance of the novel EBV mutational signatures.\\u003c/p\\u003e\"},{\"header\":\"Materials and Methods\",\"content\":\"\\u003ch2\\u003eEthics\\u003c/h2\\u003e\\n\\u003cp\\u003eThe study received approval from the Oxford Tropical Research Ethics Committee (OxTREC ref: 15-19) in the UK, the National Institute of Medical Research (NIMR/HQ/R.8a/Vol.IX/3408) in Tanzania, the National Council for Science and Technology (HS529ES) in Uganda, and the St Mary\\u0026rsquo;s Hospital Lacor Institutional Research Ethics Committee in Gulu, Uganda. Participants provided informed written consent, and/or assent for minors aged between 7 and 17 years. All study protocols adhered to the Declaration of Helsinki and international data protection regulations.\\u003c/p\\u003e\\n\\u003ch2\\u003eStudy participants and Samples\\u003c/h2\\u003e\\n\\u003cp\\u003eParticipants in the current study were enrolled in two phases. Phase I (from 2020 to 2024) involved a hospital-based prospective case-control study, originally designed to evaluate the clinical utility of liquid biopsies for diagnosing EBV-driven lymphomas, known as the AI-REAL study. The cases, which included suspected lymphoma patients, were identified from the outpatient and inpatient departments of four tertiary oncology centres in East Africa: St Mary\\u0026rsquo;s Hospital Lacor in Uganda, Muhimbili National Hospital in Dar es Salaam, Kilimanjaro Christian Medical Centre in Moshi, and Bugando Medical Centre in Mwanza, Tanzania. Patients who consented were enrolled and investigated for Burkitt lymphoma (BL) using a liquid biopsy in conjunction with conventional histopathology. The aim was to compare the accuracy of liquid biopsy with traditional histopathology. The liquid biopsy test panel included well-characterised lymphoma targets, markers, and EBV genes. BL patients with sufficient EBV levels (\\u0026ge; 2 copies per cell) were selected for whole-genome sequencing of EBV.\\u003c/p\\u003e\\n\\u003cp\\u003eIn Phase II (2024), we conducted a population-based survey of children matched to the BL cases by age, sex, and geography. Our field teams were deployed in five BL-burdened districts in northern Uganda, specifically in lower health facilities close to the villages where the BL cases originated and recruited two controls per historical BL case. Peripheral venous blood samples were collected in Qiagen Paxgene blood ccfDNA tubes (Cat no. 768165) containing a preservative that stabilises blood cells and prevents cellular DNA contamination of the plasma. The samples were transported in secure cold boxes maintained at temperatures between 2 \\u0026deg;C and 8 \\u0026deg;C to St. Mary\\u0026rsquo;s Hospital, Lacor laboratory. At the hospital, samples were separated, initially at a lower speed of 1600 \\u0026times; g for 10 minutes, and the resulting plasma was then centrifuged at a higher speed (4500 \\u0026times; g) for an additional 10 minutes. The double-spun plasma was frozen at -80 degrees until it was required for cfDNA extraction. Informed consent and/or assent from minors aged between 7 and 18 years was obtained from each participant prior to undertaking any study-related procedures. Both studies were approved by the hospital\\u0026apos;s institutional research ethics committee and the national research regulator, UNCST.\\u003c/p\\u003e\\n\\u003ch2\\u003eSample processing and EBV whole genome sequencing\\u003c/h2\\u003e\\n\\u003cp\\u003ePlasma samples were thawed, and 4-5 mL was used to extract cell-free DNA with the QIAamp Circulating Nucleic Acid Kit (Qiagen). Samples from healthy children were initially evaluated for EBV using quantitative polymerase chain reaction (qPCR). The kit (catalogue no. A58429) from Thermo Fisher Scientific amplifies the \\u003cem\\u003eEBNA1\\u003c/em\\u003e region of the EBV genome via TaqMan probes with fluorophore-based detection. EBV copies per mL were calculated from the Ct values. A total of 78 healthy control samples were assessed, but only 13 showed EBV copies sufficient for downstream analysis (copies \\u0026gt; 1000 per mL). These were selected for whole-genome sequencing. DNA libraries were prepared using the Thruplex Tag-Seq kit (Takara Bio), which involved repairing the DNA ends, ligating adapters, and amplifying the library through 7-9 PCR cycles, depending on the input concentration. The resulting libraries were purified and normalised prior to EBV capture and enrichment using the IDT xGen Hybridisation protocol (IDT). The EBV probes were designed with a 60-bp overlap, spanning 120 bp to cover the entire EBV genome. The final library was purified and normalised prior to sequencing on the Illumina MiSeq platform.\\u003c/p\\u003e\\n\\u003ch2\\u003eSelection of public EBV genomes\\u003c/h2\\u003e\\n\\u003cp\\u003ePaired-end FASTQ files from primary studies reporting EBV whole genome sequencing were retrieved from the Sequence Read Archive (SRA) under BioProject IDs PRJNA552587, PRJNA522388 and PRJNA1063319, respectively. A total of 520 paired-end FASTQ files were downloaded and filtered, with only clinical isolates considered for further analysis.\\u003c/p\\u003e\\n\\u003ch2\\u003eVariant calling\\u003c/h2\\u003e\\n\\u003cp\\u003eFASTQs from this study and those downloaded were processed collectively using a custom bioinformatics pipeline. This process involved trimming adapter sequences and removing low-quality bases (Phred score \\u0026gt; 20), followed by aligning the reads to the EBV reference genomes (NC_007605.1 for type 1, NC_009334.1 for type 2), as well as a custom hybrid reference (NC_007605.1 combined with \\u003cem\\u003eEBNA2, EBNA3s\\u003c/em\\u003e contigs from NC_009334.1) using the BWA-MEM alignment method\\u003csup\\u003e60\\u003c/sup\\u003e. The aligned reads were sorted and indexed for further analysis. Variant calling was conducted on the aligned BAM files with three different tools: VarScan\\u003csup\\u003e61\\u003c/sup\\u003e, VarDict\\u003csup\\u003e62\\u003c/sup\\u003e, and Mutect2 \\u003csup\\u003e63\\u003c/sup\\u003e. Functional annotation was carried out with SNPEff version 5.1d\\u003csup\\u003e64\\u003c/sup\\u003e, using annotation databases for both type 1 (NC_007605.1) and type 2 (NC_009334.1) reference genomes. The final VCF files listed variants identified by at least two callers and filtered for a mapping quality score of 60 or higher.\\u003c/p\\u003e\\n\\u003ch2\\u003eGenome assembly\\u003c/h2\\u003e\\n\\u003cp\\u003eGenome assembly was performed using SPAdes v4.2.0, which constructs a de Bruijn graph by decomposing sequencing reads into k-mers of multiple sizes to assemble contigs\\u003csup\\u003e65\\u003c/sup\\u003e. In our dataset, the average read length was 143 bp, and the default k-mer sizes used by SPAdes were 21, 33, and 55. The assembly quality was assessed with QUAST v5.3.0, a genome quality evaluation tool\\u003csup\\u003e66\\u003c/sup\\u003e. A composite score was calculated based on four QUAST parameters: genome fraction (GF%) \\u003cu\\u003e\\u0026gt;\\u003c/u\\u003e 70%, misassemblies \\u003cu\\u003e\\u0026lt;\\u003c/u\\u003e 1, mismatches per 100 kb \\u0026lt; 1000, and GC content within 2% of the reference, as well as a duplication ratio \\u0026le;1. Contigs achieving an overall score of 50% or higher were retained for further analysis. To order contigs against the relevant reference genome, a BLAST database was created using \\u003cem\\u003eEBNA-2\\u003c/em\\u003e EBV types 1 and 2 contigs, and the contigs were aligned to each reference using blastn, part of the NCBI BLAST+suite \\u003csup\\u003e67\\u003c/sup\\u003e. Top matching hits were assigned a strain call, and the resulting \\u003cem\\u003eEBNA 2\\u003c/em\\u003e matches were used to group the contigs into types 1 and 2. This categorisation was crucial for the public genomes, which lacked sample metadata and EBV strain information. Subsequently, the assembled contigs were ordered and oriented according to the reference genome using ABACAS (Algorithm-Based Automatic Contiguation of Assembled Sequences) \\u003csup\\u003e68\\u003c/sup\\u003e, resulting in a single pseudomolecule FASTA sequence. The assembly quality, including genome coverage, alignment rate, mean depth, and average base quality of aligned reads, was evaluated with minimap2 v2.29-r1283\\u003csup\\u003e69\\u003c/sup\\u003e. Contigs were trimmed to eliminate leading and trailing ambiguous bases, along with other sequencing artefacts, using trimAl v1.4 rev 15 \\u003csup\\u003e70\\u003c/sup\\u003e. Only assemblies with at least 60% coverage of the EBV reference genome and a mean mapping quality of at least 30 were selected for phylogenetic analysis.\\u003c/p\\u003e\\n\\u003ch2\\u003eEvaluating EBV genome diversity and population structure\\u003c/h2\\u003e\\n\\u003cp\\u003eTo infer the evolutionary and epidemiological patterns of EBV across various human populations and disease phenotypes, we aligned the FASTA genome assemblies from this study with public genomes using the multiple sequence alignment tool MAFFT v7.520\\u003csup\\u003e71\\u003c/sup\\u003e. A total of 410 genomes were aligned, including the EBV reference and the outgroup, Macacine herpesvirus 4 (NC_006146.1). The multiple sequence alignment file was trimmed using trimAl\\u003csup\\u003e70\\u003c/sup\\u003e, and known EBV repeat regions were masked using RepeatMasker version 4.1.9 in default mode\\u003csup\\u003e72\\u003c/sup\\u003e. The maximum likelihood of the phylogenetic tree was then inferred using the Randomised Accelerated Maximum Likelihood (RAxML-NG v.1.2.2) tool with the General Time Reversible (GTR-G) model\\u003csup\\u003e73\\u003c/sup\\u003e. We assessed branch support using the Transfer Bootstrap Expectation (TBE) approach with 50 replicates. Following Lemoine et al. \\u003csup\\u003e74\\u003c/sup\\u003e, we considered nodes with TBE \\u003cu\\u003e\\u0026gt;\\u003c/u\\u003e 70% to have strong support. The phylogenetic tree was visualised and annotated using R packages, ape for tree import and manipulation \\u003csup\\u003e75\\u003c/sup\\u003e, including re-rooting using the Macacine herpesvirus-4 as an outgroup, ggtree for tree visualisation \\u003csup\\u003e76\\u003c/sup\\u003e, ggtreeExtra for adding metadata annotations\\u003csup\\u003e77\\u003c/sup\\u003e, and tidyverse for data wrangling and factor reordering\\u003csup\\u003e78\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003eFor the linkage disequilibrium (LD) and principal component analysis (PCA), we analysed variants from 410 clinical samples. The variant caller files were merged and filtered for poor-quality variants (MAPQ \\u0026lt; 60) and read depth (DP) \\u0026lt; 10x. We then split the multiallelic sites into biallelic sites, left-aligned the indels using bcftools\\u003csup\\u003e79\\u003c/sup\\u003e and converted the file into PLINK binary format with genotype calls. Prior to linkage disequilibrium pruning, we estimated the average variant density per kilobase of the EBV genome across all samples and used this estimate to set the pruning window size. Variants were filtered by allele frequency (minor genotype frequency \\u0026gt; 5%) and LD. We applied a less stringent LD pruning threshold (r\\u0026sup2; \\u0026gt; 0.6, within a 1000-bp sliding window) to retain a sufficient number of informative variants for PCA. A stricter threshold (e.g., r\\u0026sup2; \\u0026lt; 0.2) would remove many variants and risk underrepresenting population structure in compact viral genomes like EBV ~ 172 kb. In total, 776 variants were included in the PCA. All analysis was conducted in PLINK v1.9 \\u003csup\\u003e80\\u003c/sup\\u003e, a whole-genome analysis toolset. The PCA was visualised in R using tidyverse, patchwork, scale and RColorBrewer packages\\u003csup\\u003e81\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003ch2\\u003eRecombination analysis via Linkage disequilibrium\\u003c/h2\\u003e\\n\\u003cp\\u003eFirst, we estimated recombination frequency in EBV genomes derived from BL and compared it to other phenotypes using linkage disequilibrium (LD) analysis. LD is a standard method to infer recombination and has been widely applied to viral genomes \\u003csup\\u003e48,82\\u003c/sup\\u003e. A steeper LD decay indicates frequent recombination events, while a slower decay reflects low recombination, preserving genetic associations over longer distances\\u003csup\\u003e48\\u003c/sup\\u003e. We extracted phenotype group-specific single-nucleotide variants (SNVs) filtered by minor allele frequency (MAF \\u0026gt; 0.01) and calculated pairwise LD (r\\u0026sup2; \\u0026gt; 0.1) within 10-kb sliding windows. The choice of a stricter LD threshold (r\\u0026sup2; \\u0026gt; 0.1) was to retain only weakly correlated variants and remove strongly linked sites. This approach has been shown to reduce redundant sites and improve the resolution of LD decay patterns, making subtle recombination signals detectable that strong LD would otherwise mask\\u003csup\\u003e83\\u003c/sup\\u003e. LD decay (r\\u0026sup2;) with genomic distance was modelled using non-linear least squares (NLS), and trends were visualised with LOESS smoothing and 95% confidence intervals. Pairwise correlations were computed using PLINK v1.9\\u003csup\\u003e80\\u003c/sup\\u003e, with statistical analyses and plotting performed in R\\u003csup\\u003e81\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eVerification of recombination and gene-level quantification studies\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eAs a sanity check, we used the rapid recombination detection programme (RDP4)\\u003csup\\u003e36\\u003c/sup\\u003e. This toolkit employs an ensemble of methods: phylogeny-based (RDP and Bootscan) and statistical models (GENECONV, MaxChi, Chimaera, 3Seq, and SiSscan) to identify recombination events and breakpoints in DNA or protein sequences. EBV genomes from 141 BL cases and 78 EBV-positive healthy controls were analysed. Multiple sequence alignments were generated using MAFFT\\u003csup\\u003e71\\u003c/sup\\u003e and trimmed using the Gblocks software\\u003csup\\u003e84\\u003c/sup\\u003e before loading into the RDP4 programme, with a default window and step sizes of 200 and 20, respectively. The output from RDP4, which includes recombination events, breakpoint positions, recombinant sequences, and major and minor parental sequences, was exported in CSV format for further analysis (Supplemental Table 2). Putative recombination events were only considered if at least five methods, including at least one phylogenetic-based method, detected the event with a threshold p-value \\u0026lt; 0.05 after Bonferroni correction. The phylogenetic trees, with and without recombinant regions, were inferred using the maximum likelihood (ML) method implemented in RAxML, employing the GTR model and 1000 bootstrap replicates\\u003csup\\u003e73\\u003c/sup\\u003e. The ML tree topologies were compared using the normalised Robinson-Foulds (RF)\\u003csup\\u003e85\\u003c/sup\\u003e implemented in the phangorn package in R\\u003csup\\u003e86\\u003c/sup\\u003e. The RF metric quantifies the proportion of differing bipartitions between two trees with values ranging from 0 (identical trees) to 1 (completely distinct topologies).\\u003c/p\\u003e\\n\\u003cp\\u003eTo assess the impact of recombination on EBV protein-coding sequences, we extracted gene-specific coding sequences (CDS) from the multiple sequence alignment files of the EBV genomes using a custom Python pipeline. The alignment file was read with Biopython (v1.85)\\u003csup\\u003e87\\u003c/sup\\u003e, and the reference genome (NC_007605.1) was used to map alignment coordinates to genomic positions. Gene coordinates from a BED annotation file were grouped by gene, and exons were merged to generate contiguous CDS sequences, accounting for wraparound genes (\\u003cem\\u003eLMP-2A\\u003c/em\\u003e). For each gene, concatenated CDS alignments were exported as FASTA files using pandas (v2.3.0)\\u003csup\\u003e88\\u003c/sup\\u003e for BED parsing and defaultdict structures for sequence assembly.\\u003c/p\\u003e\\n\\u003cp\\u003eGene-based phylogenetic trees were inferred independently for each alignment using IQ-TREE v3.0.1\\u003csup\\u003e89\\u003c/sup\\u003e with the default GTR+G substitution model and bootstrapping. The trees and FASTA alignments were then used to estimate recombination parameters in ClonalFrameML v1.13\\u003csup\\u003e90\\u003c/sup\\u003e. The choice of ClonalFrameML was based on its core methodology, which distinguishes between recombinant events and mutations on a phylogenetic tree, thereby accounting for the latter when quantifying the impact of recombination on EBV evolution as previously described\\u003csup\\u003e23\\u003c/sup\\u003e. The per-site recombination rate (⍴/bp) was then derived from the relative rate of recombination to mutation (⍴/𝛳) and average DNA length imported via recombination (\\u0026part;). The gene-specific 95% highest posterior density intervals (HPDI) were derived by propagating uncertainty from the posterior distributions of ⍴/𝛳 and \\u0026part;. Genome-wide 95% HDPI for ⍴/bp were also computed for each group. Visualisation was implemented in R using ggplot, with genes ordered by their per-site recombination rate (⍴/bp) and recombination hotspots defined as genes with per-site recombination rate exceeding the upper bound of the genome-wide HPDI.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAssessing EBV mutation frequency in BL compared to other phenotypes\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eNext, we evaluated the impact of point mutations and small indels(\\u0026lt;50bp) using the original dataset (410 genomes, including the public dataset). We assembled a comprehensive table of all variants for each phenotype. We then assigned the variants to 1-kb sliding windows based on their genomic positions. We summed the variants within each window for each sample to determine the average variant density per window and visualised this as a bar plot to identify regions with higher or lower mutation frequencies. R packages dplyr, tidyr, and ggplot were utilised for data manipulation and plotting\\u003csup\\u003e81\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003eTo identify frequently mutated genes within each phenotype, we calculated the ratio of observed to expected mutations per gene. Observed mutation counts were summed per gene and phenotype, while expected counts were estimated by multiplying the genome-wide mutation rate (average variants per sample divided by genome length in kb) by each gene\\u0026rsquo;s length in kb. For each gene-phenotype pair, we constructed a 2x2 contingency table, comparing mutations in the gene versus all other genes and performed Fisher\\u0026rsquo;s exact tests to calculate enrichment p-values. Log2fold enrichment (log2FE) scores were computed for each gene-phenotype pair with a pseudo count of 0.5 added to avoid division by zero. P values were adjusted for multiple testing using the Benjamini\\u0026ndash;Hochberg false discovery rate (FDR) method\\u003csup\\u003e91\\u003c/sup\\u003e within each phenotype. Volcano plots were used to visualise log2fold enrichment scores and FDR-adjusted significant values. All analyses were performed in R v4.4.1 using the tidyverse, ggrepel, ggplot2 and patchwork libraries\\u003csup\\u003e81\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003eTo gain a deeper understanding of EBV mutation frequency within functional gene groups, we aggregated mutation counts by gene function and compared them between BL and healthy control samples. For each gene category, group differences were assessed using the Wilcoxon rank-sum test with FDR correction for multiple testing. Cliff\\u0026rsquo;s Delta\\u003csup\\u003e92\\u003c/sup\\u003e was calculated to estimate the effect size and direction of the difference between the BL and healthy groups. The distribution of observed mutation counts per category was visualised using violin plots with overlaid boxplots.\\u003c/p\\u003e\\n\\u003ch2\\u003eEBV genome-wide association analysis and the risk of Burkitt lymphoma\\u003c/h2\\u003e\\n\\u003cp\\u003eTo identify genomic loci that might be associated with BL risk, we performed a genome-wide logistic regression analysis of EBV genomes from 141 BL cases and 78 healthy carriers. This approach is commonly employed in human genetic population studies to explore the genetic basis of complex diseases. Its adaptation for viral genomic research, known as viral genome-wide association studies (vGWAS), seeks to elucidate viral genetic factors that contribute to disease development or intricate interactions between the host and the virus by surveying the entire viral genome\\u003csup\\u003e93\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003eFirst, we merged individual sample variant caller files (VCF), split multiallelic sites introduced by VCF merging into biallelic sites using bcftools\\u003csup\\u003e79\\u003c/sup\\u003e. We then filtered out low-impact variants and retained high-confidence non-synonymous variants in both BL and healthy carriers. We merged this with sample metadata, which included information on sample origin, phenotype, strain, and principal components (PC1 and PC2) derived from whole-genome sequence variation to account for virus population structure. We used logistic regression with BL status (BL = 1, non-BL = 0) as the outcome and mutation presence as the binary variable indicating the presence of the alternate allele at a given locus. We adjusted the results for region, country, strain, and the first two principal components (PC1 and PC2) as potential confounding covariates. Logistic regression was performed using the glm function with the binomial family in R. Raw p-values for each variant were adjusted using the Benjamini-Hochberg false discovery rate (FDR) method to account for multiple comparisons\\u003csup\\u003e91\\u003c/sup\\u003e. Odds ratios (OR) were derived by exponentiating the regression coefficients, and FDR-adjusted p-values ranked variants to identify the most statistically significant associations. The Manhattan plot was used to visualise significant associations in R.\\u003c/p\\u003e\\n\\u003ch3\\u003eDe novo extraction of EBV mutation signatures and analysis\\u003c/h3\\u003e\\n\\u003cp\\u003eTo decompose the mutational signatures, we used a non-negative matrix factorisation method, a technique frequently employed in cancer genomics research\\u003csup\\u003e30,94\\u003c/sup\\u003e. We developed a Python pipeline to generate a 96-channel trinucleotide mutation matrix from VCF files. Single-nucleotide variants (SNVs) were annotated with their trinucleotide context from the EBV reference genome (NC_007605.1) using Biopython v1.85\\u003csup\\u003e87\\u003c/sup\\u003e. Mutations were normalised to the pyrimidine context, expanding the six base substitution types (C\\u0026gt;A, C\\u0026gt;G, C\\u0026gt;T, T\\u0026gt;A, T\\u0026gt;C, T\\u0026gt;G) into 96 trinucleotide contexts following the COSMIC convention\\u003csup\\u003e30\\u003c/sup\\u003e. Each sample\\u0026rsquo;s VCF file was parsed with cyvcf2 v0.30.28\\u003csup\\u003e95\\u003c/sup\\u003e, and mutation counts were aggregated into a sample-by-context matrix, which was exported for downstream signature analysis.\\u003c/p\\u003e\\n\\u003cp\\u003eWe extracted the EBV mutation signatures using MutSignature\\u003csup\\u003e94\\u003c/sup\\u003e, an R package with advanced functions for importing DNA variants, computing mutation types, and extracting mutation types via non-negative matrix factorisation (NMF)\\u003csup\\u003e96\\u003c/sup\\u003e. The software offers a wide range of compatibility, including the analysis of non-human genomes with support for the analysis of non-standard mutation types, such as tetranucleotide mutation types\\u003csup\\u003e94\\u003c/sup\\u003e. A rank survey was conducted for k = 2 to 5 across 50 iterations using both the original and randomised mutation matrices to assess signature robustness. The optimal number of signatures (k = 5) was selected based on cophenetic correlation, silhouette width, consensus clustering stability, and residual sum of squares (Suppl. Figs. 10-12). Signature exposure proportions were normalised per sample and integrated with metadata for statistical analysis. The normality of signature exposures was assessed using the Shapiro-Wilk test\\u003csup\\u003e97\\u003c/sup\\u003e (Suppl. Fig. 13). As the data exhibited non-normality, subsequent group comparisons employed beta regression models, with PC1, PC2, and strain as fixed effects to account for virus population structure and strain variation. Significant phenotype-SBS associations (FDR \\u0026lt; 0.05) were reported, along with their corresponding effect estimates. We then compared the de novo EBV signatures to the well-curated COSMIC v3.2 reference SBS signatures database using cosine similarity, calculated with the lsa R package. Cosine similarity values of 0.90 or higher were considered strong matches \\u003csup\\u003e30\\u003c/sup\\u003e. Bar plots and heatmaps were used to visualise the best matches, following the steps in previous studies \\u003csup\\u003e29,30\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003ch2\\u003eEBV gene expression studies\\u003c/h2\\u003e\\n\\u003cp\\u003eTo explore the relationship between EBV genomic diversity and viral gene expression, we analysed RNA from formalin-fixed paraffin-embedded (FFPE) tissue of 15 BL cases with available whole-genome sequencing data (See supplementary materials for detailed methods). Briefly, total RNA was extracted, rRNA-depleted, and used to generate Illumina-compatible libraries following standard protocols for degraded FFPE RNA. Libraries were multiplexed and sequenced on an Illumina NovaSeq XPlus platform (2 \\u0026times; 150 bp paired-end). Raw reads were demultiplexed with bcl2fastq. Paired-end RNA-Seq reads were first trimmed for adapters and low-quality bases using standard Illumina trimming tools. Transcript-level quantification was performed using Kallisto (v0.51.1)\\u003csup\\u003e98\\u003c/sup\\u003e. Reads were pseudo-aligned to the hybrid human-EBV transcriptome index to generate abundance estimates. All samples\\u0026rsquo; transcript-level abundance from Kallisto were aggregated to gene-level counts and imported using the tximport function in R together with the gene annotation. Differential gene expression (DGE) analysis was performed in DESeq2 (v1.48.1)\\u003csup\\u003e99\\u003c/sup\\u003e. Normalised counts were used to generate heatmaps with ComplexHeatmap (v2.24.1)\\u003csup\\u003e100\\u003c/sup\\u003e\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cspan\\u003eParents or legal guardians of children aged 3-17 years provided written informed consent, in addition to the child\\u0026apos;s assent. Participants over 17 years old gave written informed consent before any study-related procedures were carried out.\\u0026nbsp;\\u003c/span\\u003e\\u003c/p\\u003e\\u003cp\\u003e\\u003cstrong\\u003eAcknowledgements\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eWe would like to thank the patients, parents, and all study participants for their invaluable contribution to this study. We are also grateful to the study teams and our collaborators across East Africa. Our special appreciation goes to Dr. Claire El Mouden, AI-REAL Programme Manager; Dr. Faraja Chiwanga, Head of Teaching and Research at Muhimbili National Hospital. We extend our gratitude to Dr. Simon Engledow at Azenta Life Sciences for preparing the FFPE RNA libraries and conducting RNA sequencing. We also thank Ms. Gagandeep Kaur Bath for coordinating the procurement and logistics necessary for the RNA sequencing. Lastly, we thank Dr. Richard Mangwi Ayiasi at Muni University for his role in coordinating the recruitment of healthy participants for the study.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAuthor contributions\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eAS, MDO, CC, EM, and HMM conceived the AI-REAL study. IDL designed and conducted the experiments, analysed and interpreted data and wrote the first draft of the manuscript. \\u0026nbsp;IDL, DV, SH and KR developed the bioinformatic pipelines, processed and analysed the whole-genome and transcriptome datasets. AB supervised laboratory work and provided technical advice on laboratory protocol development. \\u0026nbsp;DJ, HD, and AB designed the EBV panels for both targeted and whole-genome sequencing. HN, IO, DM, PM, EM, and PSN conducted the study recruitment and pathology specimen processing. \\u0026nbsp;JS, AY, EJ, and HC performed whole-genome sequencing. \\u0026nbsp;CA, LM, AM, and EME reviewed the pathology diagnosis of all cases recruited in the study. WFM, CC, HMM, LC, EM, and MDO were responsible for administering the study sites. KR, CSKL and AS supervised the experiments, provided technical advice on experimental design, and critically revised the manuscript. All authors read and approved the final manuscript.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eCompeting interests\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eAS receives honoraria from Gilead, AbbVie, Roche, Janssen, and Illumina; unrestricted educational grants from Janssen and Gilead; and in-kind contributions from Illumina and Oxford Nanopore Technologies Ltd.\\u003c/p\\u003e\\n\\u003cp\\u003eAll other authors declare no conflict of interest.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eFunding\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis work was funded by the NIHR (NIHR-RIGHT award 200133) through UK International Development funding from the UK Government to support Global Health Research and a Ph.D. studentship from the Commonwealth Scholarship Commission. The views expressed in this publication are those of the authors and do not necessarily reflect the views of the sponsors or funders.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eData availability\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eExtended Data Fig.1\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eExtended Data Fig.2\\u003c/p\\u003e\\n\\u003cp\\u003eExtended Data Fig.3\\u003c/p\\u003e\\n\\u003cp\\u003eSupplementary Data 1 Information on analysis samples\\u003c/p\\u003e\\n\\u003cp\\u003eSupplementary Data 2 Log2fold mutation enrichment in all samples\\u003c/p\\u003e\\n\\u003cp\\u003eRaw sequencing data (FASTQs) and novel EBV genomes will be uploaded to the appropriate biorepositories.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAdditional information\\u0026nbsp;\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eFurther information on study design, methods (including detailed testing protocols), statistical tables, and figures is available in the supplementary information.\\u0026nbsp;\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n\\u003cli\\u003eVetsika, E.-K. \\u0026amp; Callan, M. Infectious mononucleosis and Epstein-Barr virus. \\u003cem\\u003eExpert reviews in molecular medicine\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, 1-16 (2004).\\u003c/li\\u003e\\n\\u003cli\\u003eDunmire, S.K., Verghese, P.S. \\u0026amp; Balfour, H.H. Primary Epstein-Barr virus infection. \\u003cem\\u003eJournal of Clinical Virology\\u003c/em\\u003e \\u003cstrong\\u003e102\\u003c/strong\\u003e, 84-92 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eM\\u0026uuml;nz, C. Latency and lytic replication in Epstein\\u0026ndash;Barr virus-associated oncogenesis. \\u003cem\\u003eNature Reviews Microbiology\\u003c/em\\u003e \\u003cstrong\\u003e17\\u003c/strong\\u003e, 691-700 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eWong, Y., Meehan, M.T., Burrows, S.R., Doolan, D.L. \\u0026amp; Miles, J.J. Estimating the global burden of Epstein-Barr virus-related cancers. \\u003cem\\u003eJ Cancer Res Clin Oncol\\u003c/em\\u003e \\u003cstrong\\u003e148\\u003c/strong\\u003e, 31-46 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eBjornevik, K., M\\u0026uuml;nz, C., Cohen, J.I. \\u0026amp; Ascherio, A. Epstein-Barr virus as a leading cause of multiple sclerosis: mechanisms and implications. \\u003cem\\u003eNat Rev Neurol\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 160-171 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eHouen, G. \\u0026amp; Trier, N.H. Epstein-Barr Virus and Systemic Autoimmune Diseases. \\u003cem\\u003eFrontiers in Immunology\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e(2021).\\u003c/li\\u003e\\n\\u003cli\\u003eH\\u0026auml;mmerl, L., Colombet, M., Rochford, R., Ogwang, D.M. \\u0026amp; Parkin, D.M. The burden of Burkitt lymphoma in Africa. \\u003cem\\u003eInfect Agent Cancer\\u003c/em\\u003e \\u003cstrong\\u003e14\\u003c/strong\\u003e, 17 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eYu, H.\\u003cem\\u003e et al.\\u003c/em\\u003e The global burden of nasopharyngeal carcinoma from 2009 to 2019: an observational study based on the Global Burden of Disease Study 2019. \\u003cem\\u003eEuropean Archives of Oto-Rhino-Laryngology\\u003c/em\\u003e \\u003cstrong\\u003e279\\u003c/strong\\u003e, 1519-1533 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eKirimunda, S.\\u003cem\\u003e et al.\\u003c/em\\u003e Variation in the Human Leukocyte Antigen system and risk for endemic Burkitt lymphoma in northern Uganda. \\u003cem\\u003eBritish Journal of Haematology\\u003c/em\\u003e \\u003cstrong\\u003e189\\u003c/strong\\u003e, 489-499 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eKanda, T., Yajima, M. \\u0026amp; Ikuta, K. Epstein‐Barr virus strain variation and cancer. \\u003cem\\u003eCancer science\\u003c/em\\u003e \\u003cstrong\\u003e110\\u003c/strong\\u003e, 1132-1139 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eBurkitt, D.P. Etiology of Burkitt\\u0026apos;s lymphoma\\u0026mdash;an alternative hypothesis to a vectored virus. \\u003cem\\u003eJournal of the National Cancer Institute\\u003c/em\\u003e \\u003cstrong\\u003e42\\u003c/strong\\u003e, 19-28 (1969).\\u003c/li\\u003e\\n\\u003cli\\u003eBahri, R.\\u003cem\\u003e et al.\\u003c/em\\u003e Epstein-Barr Virus (EBV) is mostly latent and clonal in angioimmunoblastic T cell lymphoma (AITL). \\u003cem\\u003eCancers\\u003c/em\\u003e \\u003cstrong\\u003e14\\u003c/strong\\u003e, 2899 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003ePathmanathan, R., Prasad, U., Sadler, R., Flynn, K. \\u0026amp; Raab-Traub, N. Clonal proliferations of cells infected with Epstein\\u0026ndash;Barr virus in preinvasive lesions related to nasopharyngeal carcinoma. \\u003cem\\u003eNew England Journal of Medicine\\u003c/em\\u003e \\u003cstrong\\u003e333\\u003c/strong\\u003e, 693-698 (1995).\\u003c/li\\u003e\\n\\u003cli\\u003eYoung, L.S., Arrand, J.R. \\u0026amp; Murray, P.G. EBV gene expression and regulation. in \\u003cem\\u003eHuman Herpesviruses: Biology, Therapy, and Immunoprophylaxis\\u003c/em\\u003e (eds. Arvin, A.\\u003cem\\u003e et al.\\u003c/em\\u003e) (Cambridge University Press, Cambridge, 2007).\\u003c/li\\u003e\\n\\u003cli\\u003eMurata, T.\\u003cem\\u003e et al.\\u003c/em\\u003e Molecular basis of epstein\\u0026ndash;barr virus latency establishment and lytic reactivation. \\u003cem\\u003eViruses\\u003c/em\\u003e \\u003cstrong\\u003e13\\u003c/strong\\u003e, 2344 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eInagaki, T.\\u003cem\\u003e et al.\\u003c/em\\u003e Direct evidence of abortive lytic infection-mediated establishment of Epstein-Barr virus latency during B-cell infection. \\u003cem\\u003eFrontiers in Microbiology\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 575255 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eXue, S.a.\\u003cem\\u003e et al.\\u003c/em\\u003e Promiscuous expression of Epstein‐Barr virus genes in Burkitt\\u0026apos;s lymphoma from the central African country Malawi. \\u003cem\\u003eInternational journal of cancer\\u003c/em\\u003e \\u003cstrong\\u003e99\\u003c/strong\\u003e, 635-643 (2002).\\u003c/li\\u003e\\n\\u003cli\\u003eAbate, F.\\u003cem\\u003e et al.\\u003c/em\\u003e Distinct Viral and Mutational Spectrum of Endemic Burkitt Lymphoma. \\u003cem\\u003ePLOS Pathogens\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, e1005158 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eAgwati, E.O.\\u003cem\\u003e et al.\\u003c/em\\u003e Profiling genome-wide recombination in Epstein Barr virus reveals type-specific patterns and associations with endemic-Burkitt lymphoma. \\u003cem\\u003eVirology Journal\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 208 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eYao, Q.\\u003cem\\u003e et al.\\u003c/em\\u003e Isolation of intertypic recombinants of Epstein-Barr virus from T-cell-immunocompromised individuals. \\u003cem\\u003eJournal of virology\\u003c/em\\u003e \\u003cstrong\\u003e70\\u003c/strong\\u003e, 4895-4903 (1996).\\u003c/li\\u003e\\n\\u003cli\\u003eP\\u0026eacute;rez-Losada, M., Arenas, M., Gal\\u0026aacute;n, J.C., Palero, F. \\u0026amp; Gonz\\u0026aacute;lez-Candelas, F. Recombination in viruses: mechanisms, methods of study, and evolutionary consequences. \\u003cem\\u003eInfect Genet Evol\\u003c/em\\u003e \\u003cstrong\\u003e30\\u003c/strong\\u003e, 296-307 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eBerenstein, A.J., Lorenzetti, M.A. \\u0026amp; Preciado, M.V. Recombination rates along the entire Epstein Barr virus genome display a highly heterogeneous landscape. \\u003cem\\u003eInfection, Genetics and Evolution\\u003c/em\\u003e \\u003cstrong\\u003e65\\u003c/strong\\u003e, 96-103 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eZanella, L.\\u003cem\\u003e et al.\\u003c/em\\u003e A reliable Epstein-Barr Virus classification based on phylogenomic and population analyses. \\u003cem\\u003eScientific Reports\\u003c/em\\u003e \\u003cstrong\\u003e9\\u003c/strong\\u003e, 9829 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eXu, M.\\u003cem\\u003e et al.\\u003c/em\\u003e Genome sequencing analysis identifies Epstein\\u0026ndash;Barr virus subtypes associated with high risk of nasopharyngeal carcinoma. \\u003cem\\u003eNature Genetics\\u003c/em\\u003e \\u003cstrong\\u003e51\\u003c/strong\\u003e, 1131-1136 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eRickinson, A.B., Young, L.S. \\u0026amp; Rowe, M. Influence of the Epstein-Barr virus nuclear antigen EBNA 2 on the growth phenotype of virus-transformed B cells. \\u003cem\\u003eJ Virol\\u003c/em\\u003e \\u003cstrong\\u003e61\\u003c/strong\\u003e, 1310-7 (1987).\\u003c/li\\u003e\\n\\u003cli\\u003eSample, J.\\u003cem\\u003e et al.\\u003c/em\\u003e Epstein-Barr virus types 1 and 2 differ in their EBNA-3A, EBNA-3B, and EBNA-3C genes. \\u003cem\\u003eJournal of Virology\\u003c/em\\u003e \\u003cstrong\\u003e64\\u003c/strong\\u003e, 4084-4092 (1990).\\u003c/li\\u003e\\n\\u003cli\\u003eHui, K.F.\\u003cem\\u003e et al.\\u003c/em\\u003e High risk Epstein‐Barr virus variants characterized by distinct polymorphisms in the EBER locus are strongly associated with nasopharyngeal carcinoma. \\u003cem\\u003eInternational journal of cancer\\u003c/em\\u003e \\u003cstrong\\u003e144\\u003c/strong\\u003e, 3031-3042 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eda Costa, V.G., Marques-Silva, A.C. \\u0026amp; Moreli, M.L. The Epstein-Barr virus latent membrane protein-1 (LMP1) 30-bp deletion and XhoI-polymorphism in nasopharyngeal carcinoma: a meta-analysis of observational studies. \\u003cem\\u003eSystematic Reviews\\u003c/em\\u003e \\u003cstrong\\u003e4\\u003c/strong\\u003e, 46 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eAlexandrov, L.B., Nik-Zainal, S., Wedge, D.C., Campbell, P.J. \\u0026amp; Stratton, M.R. Deciphering signatures of mutational processes operative in human cancer. \\u003cem\\u003eCell reports\\u003c/em\\u003e \\u003cstrong\\u003e3\\u003c/strong\\u003e, 246-259 (2013).\\u003c/li\\u003e\\n\\u003cli\\u003eAlexandrov, L.B.\\u003cem\\u003e et al.\\u003c/em\\u003e The repertoire of mutational signatures in human cancer. \\u003cem\\u003eNature\\u003c/em\\u003e \\u003cstrong\\u003e578\\u003c/strong\\u003e, 94-101 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eGodsmark, G., Rios, L.A.D.S. \\u0026amp; Mowla, S. Activation-induced cytidine deaminase promotes proliferation and enhances chemoresistance and migration in B-cell lymphoma. \\u003cem\\u003eAnticancer Research\\u003c/em\\u003e \\u003cstrong\\u003e41\\u003c/strong\\u003e, 237-247 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eShilova, O., Tsyba, D. \\u0026amp; Shilov, E. Mutagenic activity of AID/APOBEC deaminases in antiviral defense and carcinogenesis. \\u003cem\\u003eMolecular Biology\\u003c/em\\u003e \\u003cstrong\\u003e56\\u003c/strong\\u003e, 46-58 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eWan, J.C.\\u003cem\\u003e et al.\\u003c/em\\u003e Liquid biopsies come of age: towards implementation of circulating tumour DNA. \\u003cem\\u003eNature Reviews Cancer\\u003c/em\\u003e \\u003cstrong\\u003e17\\u003c/strong\\u003e, 223-238 (2017).\\u003c/li\\u003e\\n\\u003cli\\u003eXu, M.\\u003cem\\u003e et al.\\u003c/em\\u003e Genome-wide profiling of Epstein-Barr virus integration by targeted sequencing in Epstein-Barr virus associated malignancies. \\u003cem\\u003eTheranostics\\u003c/em\\u003e \\u003cstrong\\u003e9\\u003c/strong\\u003e, 1115 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eXiao, K.\\u003cem\\u003e et al.\\u003c/em\\u003e Genome-wide analysis of Epstein-Barr virus (EBV) integration and strain in C666-1 and Raji cells. \\u003cem\\u003eJournal of Cancer\\u003c/em\\u003e \\u003cstrong\\u003e7\\u003c/strong\\u003e, 214 (2016).\\u003c/li\\u003e\\n\\u003cli\\u003eMartin, D.P., Murrell, B., Golden, M., Khoosal, A. \\u0026amp; Muhire, B. RDP4: Detection and analysis of recombination patterns in virus genomes. \\u003cem\\u003eVirus Evolution\\u003c/em\\u003e \\u003cstrong\\u003e1\\u003c/strong\\u003e(2015).\\u003c/li\\u003e\\n\\u003cli\\u003eSausen, D., Poirier, M., Spiers, L. \\u0026amp; Smith, E. Mechanisms of T cell evasion by Epstein-Barr virus and implications for tumor survival. \\u003cem\\u003eFrontiers in immunology\\u003c/em\\u003e \\u003cstrong\\u003e14\\u003c/strong\\u003e, 1289313 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eRessing, M.E.\\u003cem\\u003e et al.\\u003c/em\\u003e Epstein-Barr virus evasion of CD8+ and CD4+ T cell immunity via concerted actions of multiple gene products. in \\u003cem\\u003eSeminars in cancer biology \\u003c/em\\u003eVol. 18 397-408 (Elsevier, 2008).\\u003c/li\\u003e\\n\\u003cli\\u003eZuo, J.\\u003cem\\u003e et al.\\u003c/em\\u003e The Epstein-Barr virus G-protein-coupled receptor contributes to immune evasion by targeting MHC class I molecules for degradation. \\u003cem\\u003ePLoS pathogens\\u003c/em\\u003e \\u003cstrong\\u003e5\\u003c/strong\\u003e, e1000255 (2009).\\u003c/li\\u003e\\n\\u003cli\\u003eJochum, S., Moosmann, A., Lang, S., Hammerschmidt, W. \\u0026amp; Zeidler, R. The EBV immunoevasins vIL-10 and BNLF2a protect newly infected B cells from immune recognition and elimination. \\u003cem\\u003ePLoS pathogens\\u003c/em\\u003e \\u003cstrong\\u003e8\\u003c/strong\\u003e, e1002704 (2012).\\u003c/li\\u003e\\n\\u003cli\\u003eYap, L.F., Wong, A.K.C., Paterson, I.C. \\u0026amp; Young, L.S. Functional implications of Epstein-Barr virus lytic genes in carcinogenesis. \\u003cem\\u003eCancers\\u003c/em\\u003e \\u003cstrong\\u003e14\\u003c/strong\\u003e, 5780 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eYoung, L.S. \\u0026amp; Rickinson, A.B. Epstein\\u0026ndash;Barr virus: 40 years on. \\u003cem\\u003eNature Reviews Cancer\\u003c/em\\u003e \\u003cstrong\\u003e4\\u003c/strong\\u003e, 757-768 (2004).\\u003c/li\\u003e\\n\\u003cli\\u003eBalfour Jr, H.H.\\u003cem\\u003e et al.\\u003c/em\\u003e Age-specific prevalence of Epstein\\u0026ndash;Barr virus infection among individuals aged 6\\u0026ndash;19 years in the United States and factors affecting its acquisition. \\u003cem\\u003eThe Journal of infectious diseases\\u003c/em\\u003e \\u003cstrong\\u003e208\\u003c/strong\\u003e, 1286-1293 (2013).\\u003c/li\\u003e\\n\\u003cli\\u003eDe-Th\\u0026eacute;, G.\\u003cem\\u003e et al.\\u003c/em\\u003e Epidemiological evidence for causal relationship between Epstein-Barr virus and Burkitt\\u0026apos;s lymphoma from Ugandan prospective study. \\u003cem\\u003eNature\\u003c/em\\u003e \\u003cstrong\\u003e274\\u003c/strong\\u003e, 756-761 (1978).\\u003c/li\\u003e\\n\\u003cli\\u003eBriercheck, E.L.\\u003cem\\u003e et al.\\u003c/em\\u003e Geographic EBV variants confound disease-specific variant interpretation and predict variable immune therapy responses. \\u003cem\\u003eBlood Advances\\u003c/em\\u003e \\u003cstrong\\u003e8\\u003c/strong\\u003e, 3731-3744 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eChiara, M.\\u003cem\\u003e et al.\\u003c/em\\u003e Geographic population structure in Epstein-Barr virus revealed by comparative genomics. \\u003cem\\u003eGenome biology and evolution\\u003c/em\\u003e \\u003cstrong\\u003e8\\u003c/strong\\u003e, 3284-3291 (2016).\\u003c/li\\u003e\\n\\u003cli\\u003ePalser, A.L.\\u003cem\\u003e et al.\\u003c/em\\u003e Genome diversity of Epstein-Barr virus from multiple tumor types and normal infection. \\u003cem\\u003eJournal of virology\\u003c/em\\u003e \\u003cstrong\\u003e89\\u003c/strong\\u003e, 5222-5237 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003ePati\\u0026ntilde;o-Galindo, J.\\u0026Aacute;., Filip, I. \\u0026amp; Rabadan, R. Global patterns of recombination across human viruses. \\u003cem\\u003eMolecular biology and evolution\\u003c/em\\u003e \\u003cstrong\\u003e38\\u003c/strong\\u003e, 2520-2531 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eLevitskaya, J., Sharipo, A., Leonchiks, A., Ciechanover, A. \\u0026amp; Masucci, M.G. Inhibition of ubiquitin/proteasome-dependent protein degradation by the Gly-Ala repeat domain of the Epstein-Barr virus nuclear antigen 1. \\u003cem\\u003eProc Natl Acad Sci U S A\\u003c/em\\u003e \\u003cstrong\\u003e94\\u003c/strong\\u003e, 12616-21 (1997).\\u003c/li\\u003e\\n\\u003cli\\u003eXue, S.A. \\u0026amp; Griffin, B.E. Complexities associated with expression of Epstein-Barr virus (EBV) lytic origins of DNA replication. \\u003cem\\u003eNucleic Acids Res\\u003c/em\\u003e \\u003cstrong\\u003e35\\u003c/strong\\u003e, 3391-406 (2007).\\u003c/li\\u003e\\n\\u003cli\\u003eMancao, C. \\u0026amp; Hammerschmidt, W. Epstein-Barr virus latent membrane protein 2A is a B-cell receptor mimic and essential for B-cell survival. \\u003cem\\u003eBlood\\u003c/em\\u003e \\u003cstrong\\u003e110\\u003c/strong\\u003e, 3715-21 (2007).\\u003c/li\\u003e\\n\\u003cli\\u003eChen, X., Kamranvar, S.A. \\u0026amp; Masucci, M.G. Oxidative stress enables Epstein-Barr virus-induced B-cell transformation by posttranscriptional regulation of viral and cellular growth-promoting factors. \\u003cem\\u003eOncogene\\u003c/em\\u003e \\u003cstrong\\u003e35\\u003c/strong\\u003e, 3807-16 (2016).\\u003c/li\\u003e\\n\\u003cli\\u003eKikuchi, K.\\u003cem\\u003e et al.\\u003c/em\\u003e Overexpression of Activation-Induced Cytidine Deaminase in MTX- and Age-Related Epstein-Barr Virus-Associated B-Cell Lymphoproliferative Disorders of the Head and Neck. \\u003cem\\u003eJ Oncol\\u003c/em\\u003e \\u003cstrong\\u003e2015\\u003c/strong\\u003e, 605750 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eLirussi, L. \\u0026amp; Nilsen, H.L. DNA glycosylases define the outcome of endogenous base modifications. \\u003cem\\u003eInternational Journal of Molecular Sciences\\u003c/em\\u003e \\u003cstrong\\u003e24\\u003c/strong\\u003e, 10307 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eGunnarsson, R.\\u003cem\\u003e et al.\\u003c/em\\u003e Single base substitution and insertion/deletion mutational signatures in adult core binding factor acute myeloid leukemia. \\u003cem\\u003eLeukemia\\u003c/em\\u003e \\u003cstrong\\u003e36\\u003c/strong\\u003e, 1681-1684 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eYaacov, A.\\u003cem\\u003e et al.\\u003c/em\\u003e Cancer mutational signatures identification in clinical assays using neural embedding-based representations. \\u003cem\\u003eCell Rep Med\\u003c/em\\u003e \\u003cstrong\\u003e5\\u003c/strong\\u003e, 101608 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eKarihtala, P., Kilpivaara, O. \\u0026amp; Porvari, K. Mutational signatures and their association with survival and gene expression in urological carcinomas. \\u003cem\\u003eNeoplasia\\u003c/em\\u003e \\u003cstrong\\u003e44\\u003c/strong\\u003e, 100933 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003ePonomarev, G.V.\\u003cem\\u003e et al.\\u003c/em\\u003e APOBEC mutagenesis is low in most types of non-B DNA structures. \\u003cem\\u003eiScience\\u003c/em\\u003e \\u003cstrong\\u003e25\\u003c/strong\\u003e, 104535 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eTomkova, M., Tomek, J., Kriaucionis, S. \\u0026amp; Schuster-B\\u0026ouml;ckler, B. Mutational signature distribution varies with DNA replication timing and strand asymmetry. \\u003cem\\u003eGenome Biology\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 129 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eVasimuddin, M., Misra, S., Li, H. \\u0026amp; Aluru, S. Efficient architecture-aware acceleration of BWA-MEM for multicore systems. in \\u003cem\\u003e2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS) \\u003c/em\\u003e314-324 (IEEE, 2019).\\u003c/li\\u003e\\n\\u003cli\\u003eKoboldt, D.C.\\u003cem\\u003e et al.\\u003c/em\\u003e VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. \\u003cem\\u003eGenome research\\u003c/em\\u003e \\u003cstrong\\u003e22\\u003c/strong\\u003e, 568-576 (2012).\\u003c/li\\u003e\\n\\u003cli\\u003eLai, Z.\\u003cem\\u003e et al.\\u003c/em\\u003e VarDict: a novel and versatile variant caller for next-generation sequencing in cancer research. \\u003cem\\u003eNucleic acids research\\u003c/em\\u003e \\u003cstrong\\u003e44\\u003c/strong\\u003e, e108-e108 (2016).\\u003c/li\\u003e\\n\\u003cli\\u003eBenjamin, D.\\u003cem\\u003e et al.\\u003c/em\\u003e Calling somatic SNVs and indels with Mutect2. \\u003cem\\u003eBiorxiv\\u003c/em\\u003e, 861054 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eCingolani, P.\\u003cem\\u003e et al.\\u003c/em\\u003e A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. \\u003cem\\u003efly\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, 80-92 (2012).\\u003c/li\\u003e\\n\\u003cli\\u003ePrjibelski, A., Antipov, D., Meleshko, D., Lapidus, A. \\u0026amp; Korobeynikov, A. Using SPAdes De Novo Assembler. \\u003cem\\u003eCurrent Protocols in Bioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e70\\u003c/strong\\u003e, e102 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eGurevich, A., Saveliev, V., Vyahhi, N. \\u0026amp; Tesler, G. QUAST: quality assessment tool for genome assemblies. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e29\\u003c/strong\\u003e, 1072-1075 (2013).\\u003c/li\\u003e\\n\\u003cli\\u003eCamacho, C.\\u003cem\\u003e et al.\\u003c/em\\u003e BLAST+: architecture and applications. \\u003cem\\u003eBMC bioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e10\\u003c/strong\\u003e, 421 (2009).\\u003c/li\\u003e\\n\\u003cli\\u003eZhou, W.\\u003cem\\u003e et al.\\u003c/em\\u003e ABACUS: An Electronic Structure Analysis Package for the AI Era. \\u003cem\\u003earXiv preprint arXiv:2501.08697\\u003c/em\\u003e (2025).\\u003c/li\\u003e\\n\\u003cli\\u003eLi, H. Minimap2: pairwise alignment for nucleotide sequences. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e34\\u003c/strong\\u003e, 3094-3100 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eCapella-Guti\\u0026eacute;rrez, S., Silla-Mart\\u0026iacute;nez, J.M. \\u0026amp; Gabald\\u0026oacute;n, T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e25\\u003c/strong\\u003e, 1972-1973 (2009).\\u003c/li\\u003e\\n\\u003cli\\u003eKatoh, K., Misawa, K., Kuma, K.i. \\u0026amp; Miyata, T. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. \\u003cem\\u003eNucleic acids research\\u003c/em\\u003e \\u003cstrong\\u003e30\\u003c/strong\\u003e, 3059-3066 (2002).\\u003c/li\\u003e\\n\\u003cli\\u003eSmit, A. Smit, AFA, Hubley, R \\u0026amp; Green, P. RepeatMasker Open-3.0. (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eTogkousidis, A., Kozlov, O.M., Haag, J., H\\u0026ouml;hler, D. \\u0026amp; Stamatakis, A. Adaptive RAxML-NG: accelerating phylogenetic inference under maximum likelihood using dataset difficulty. \\u003cem\\u003eMolecular Biology and Evolution\\u003c/em\\u003e \\u003cstrong\\u003e40\\u003c/strong\\u003e, msad227 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eLemoine, F.\\u003cem\\u003e et al.\\u003c/em\\u003e Renewing Felsenstein\\u0026rsquo;s phylogenetic bootstrap in the era of big data. \\u003cem\\u003eNature\\u003c/em\\u003e \\u003cstrong\\u003e556\\u003c/strong\\u003e, 452-456 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eParadis, E. \\u0026amp; Schliep, K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e35\\u003c/strong\\u003e, 526-528 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eYu, G., Smith, D.K., Zhu, H., Guan, Y. \\u0026amp; Lam, T.T.Y. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. \\u003cem\\u003eMethods in ecology and evolution\\u003c/em\\u003e \\u003cstrong\\u003e8\\u003c/strong\\u003e, 28-36 (2017).\\u003c/li\\u003e\\n\\u003cli\\u003eXu, S.\\u003cem\\u003e et al.\\u003c/em\\u003e ggtreeExtra: compact visualization of richly annotated phylogenetic data. \\u003cem\\u003eMolecular biology and evolution\\u003c/em\\u003e \\u003cstrong\\u003e38\\u003c/strong\\u003e, 4039-4042 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eWickham, H.\\u003cem\\u003e et al.\\u003c/em\\u003e Welcome to the Tidyverse. \\u003cem\\u003eJournal of open source software\\u003c/em\\u003e \\u003cstrong\\u003e4\\u003c/strong\\u003e, 1686 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eDanecek, P.\\u003cem\\u003e et al.\\u003c/em\\u003e Twelve years of SAMtools and BCFtools. \\u003cem\\u003eGigaScience\\u003c/em\\u003e \\u003cstrong\\u003e10\\u003c/strong\\u003e(2021).\\u003c/li\\u003e\\n\\u003cli\\u003eChang, C.C.\\u003cem\\u003e et al.\\u003c/em\\u003e Second-generation PLINK: rising to the challenge of larger and richer datasets. \\u003cem\\u003eGigascience\\u003c/em\\u003e \\u003cstrong\\u003e4\\u003c/strong\\u003e, s13742-015-0047-8 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eTeam, R.C. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. \\u003cem\\u003ehttp://www\\u003c/em\\u003e\\u003cem\\u003e. R-project. org/\\u003c/em\\u003e (2016).\\u003c/li\\u003e\\n\\u003cli\\u003eRomero, E.V. \\u0026amp; Feder, A.F. Elevated HIV viral load is associated with higher recombination rate in vivo. \\u003cem\\u003eMolecular Biology and Evolution\\u003c/em\\u003e \\u003cstrong\\u003e41\\u003c/strong\\u003e, msad260 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eVos, P.G.\\u003cem\\u003e et al.\\u003c/em\\u003e Evaluation of LD decay and various LD-decay estimators in simulated and SNP-array data of tetraploid potato. \\u003cem\\u003eTheor Appl Genet\\u003c/em\\u003e \\u003cstrong\\u003e130\\u003c/strong\\u003e, 123-135 (2017).\\u003c/li\\u003e\\n\\u003cli\\u003eCastresana, J. Selection of Conserved Blocks from Multiple Alignments for Their Use in Phylogenetic Analysis. \\u003cem\\u003eMolecular Biology and Evolution\\u003c/em\\u003e \\u003cstrong\\u003e17\\u003c/strong\\u003e, 540-552 (2000).\\u003c/li\\u003e\\n\\u003cli\\u003eRobinson, D.F. \\u0026amp; Foulds, L.R. Comparison of phylogenetic trees. \\u003cem\\u003eMathematical biosciences\\u003c/em\\u003e \\u003cstrong\\u003e53\\u003c/strong\\u003e, 131-147 (1981).\\u003c/li\\u003e\\n\\u003cli\\u003eSchliep, K.P. phangorn: phylogenetic analysis in R. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e27\\u003c/strong\\u003e, 592-593 (2011).\\u003c/li\\u003e\\n\\u003cli\\u003eCock, P.J.A.\\u003cem\\u003e et al.\\u003c/em\\u003e Biopython: freely available Python tools for computational molecular biology and bioinformatics. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e25\\u003c/strong\\u003e, 1422-1423 (2009).\\u003c/li\\u003e\\n\\u003cli\\u003eTeam, T.P.D. pandas-dev/pandas: Pandas. \\u003cem\\u003eZenodo, February\\u003c/em\\u003e (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eWong, T.K.\\u003cem\\u003e et al.\\u003c/em\\u003e IQ-TREE 3: Phylogenomic Inference Software using Complex Evolutionary Models. (2025).\\u003c/li\\u003e\\n\\u003cli\\u003eDidelot, X. \\u0026amp; Wilson, D.J. ClonalFrameML: efficient inference of recombination in whole bacterial genomes. \\u003cem\\u003ePLoS computational biology\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, e1004041 (2015).\\u003c/li\\u003e\\n\\u003cli\\u003eBenjamini, Y. \\u0026amp; Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. \\u003cem\\u003eJournal of the Royal statistical society: series B (Methodological)\\u003c/em\\u003e \\u003cstrong\\u003e57\\u003c/strong\\u003e, 289-300 (1995).\\u003c/li\\u003e\\n\\u003cli\\u003eMeissel, K. \\u0026amp; Yao, E. Using Cliff\\u0026rsquo;s Delta as a Non-Parametric Effect Size Measure: An Accessible Web App and R Tutorial. 2024. (1977).\\u003c/li\\u003e\\n\\u003cli\\u003eButković, A. \\u0026amp; Elena, S.F. Genome-wide association studies of viral infections\\u0026mdash;A short guide to a successful experimental and statistical analysis. \\u003cem\\u003eFrontiers in Systems Biology\\u003c/em\\u003e \\u003cstrong\\u003e2\\u003c/strong\\u003e, 1005758 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eFantini, D., Vidimar, V., Yu, Y., Condello, S. \\u0026amp; Meeks, J.J. MutSignatures: an R package for extraction and analysis of cancer mutational signatures. \\u003cem\\u003eScientific Reports\\u003c/em\\u003e \\u003cstrong\\u003e10\\u003c/strong\\u003e, 18217 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003ePedersen, B.S. \\u0026amp; Quinlan, A.R. Who\\u0026rsquo;s who? Detecting and resolving sample anomalies in human DNA sequencing studies with peddy. \\u003cem\\u003eThe American Journal of Human Genetics\\u003c/em\\u003e \\u003cstrong\\u003e100\\u003c/strong\\u003e, 406-413 (2017).\\u003c/li\\u003e\\n\\u003cli\\u003eGaujoux, R. \\u0026amp; Seoighe, C. A flexible R package for nonnegative matrix factorization. \\u003cem\\u003eBMC bioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 367 (2010).\\u003c/li\\u003e\\n\\u003cli\\u003eShapiro, S.S. \\u0026amp; Wilk, M.B. An analysis of variance test for normality (complete samples). \\u003cem\\u003eBiometrika\\u003c/em\\u003e \\u003cstrong\\u003e52\\u003c/strong\\u003e, 591-611 (1965).\\u003c/li\\u003e\\n\\u003cli\\u003eBray, N.L., Pimentel, H., Melsted, P. \\u0026amp; Pachter, L. Near-optimal probabilistic RNA-seq quantification. \\u003cem\\u003eNature biotechnology\\u003c/em\\u003e \\u003cstrong\\u003e34\\u003c/strong\\u003e, 525-527 (2016).\\u003c/li\\u003e\\n\\u003cli\\u003eLove, M.I., Huber, W. \\u0026amp; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. \\u003cem\\u003eGenome Biology\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 550 (2014).\\u003c/li\\u003e\\n\\u003cli\\u003eGu, Z., Eils, R. \\u0026amp; Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e \\u003cstrong\\u003e32\\u003c/strong\\u003e, 2847-2849 (2016).\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"nature-portfolio\",\"isNatureJournal\":true,\"hasQc\":false,\"allowDirectSubmit\":false,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"\",\"title\":\"Nature Portfolio\",\"twitterHandle\":\"\",\"acdcEnabled\":false,\"dfaEnabled\":false,\"editorialSystem\":\"ejp\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false},\"keywords\":\"Epstein-Barr virus, latent-lytic replication, recombination, mutations, genetic variation and Burkitt lymphoma\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-7473738/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-7473738/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eEpstein\\u0026ndash;Barr virus (EBV) is a ubiquitous human herpesvirus linked to multiple malignancies, but its role in disease remains uncertain. We analysed 410 EBV genomes, including 64 newly sequenced isolates from plasma of East African children with and without Burkitt lymphoma (BL), integrating whole-genome and transcriptomic data. Population genomic analyses revealed marked regional structure with strain-level diversity shaped by both recombination and mutation, particularly within latency and immune-evasion genes (\\u003cem\\u003eEBNA-1\\u003c/em\\u003e, \\u003cem\\u003eLMP-1\\u003c/em\\u003e, \\u003cem\\u003eBCRF1\\u003c/em\\u003e, \\u003cem\\u003eBNLF2a\\u003c/em\\u003e), reflecting selective pressure from host immunity and tumour microenvironment. We identified a BL-associated mutational signature (SBS_EBV3) enriched for T\\u0026thinsp;\\u0026gt;\\u0026thinsp;C and C\\u0026thinsp;\\u0026gt;\\u0026thinsp;T substitutions in G-rich contexts, with limited similarity to known COSMIC signatures (closest match SBS54, cosine similarity 0.74). Transcriptomic profiling demonstrated a mixed latent\\u0026ndash;lytic expression programme in BL, potentially promoting recombination and mutagenesis. These findings define new features of EBV evolution in BL and highlight opportunities for diagnostics and vaccines targeting both latent and lytic antigens.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Evidence of latent-lytic replication in EBV-positive Burkitt lymphoma from whole genome and transcriptome sequencing\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-10-17 12:42:52\",\"doi\":\"10.21203/rs.3.rs-7473738/v1\",\"editorialEvents\":[],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"nature-communications\",\"isNatureJournal\":true,\"hasQc\":false,\"allowDirectSubmit\":false,\"externalIdentity\":\"NCOMMS\",\"sideBox\":\"Learn more about [Nature Communications](http://www.nature.com/ncomms/)\",\"snPcode\":\"\",\"submissionUrl\":\"https://mts-ncomms.nature.com/\",\"title\":\"Nature Communications\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"ejp\",\"reportingPortfolio\":\"Nature Communications\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":false}}],\"origin\":\"\",\"ownerIdentity\":\"5cdcb97d-cdfe-47af-bfdc-268374c0a15a\",\"owner\":[],\"postedDate\":\"October 17th, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[{\"id\":54342448,\"name\":\"Biological sciences/Cancer/Tumour virus infections\"},{\"id\":54342449,\"name\":\"Health sciences/Medical research/Genetics research\"}],\"tags\":[],\"updatedAt\":\"2025-10-17T12:42:52+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-10-17 12:42:52\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-7473738\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-7473738\",\"identity\":\"rs-7473738\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}