Whole exome-seq and RNA-seq data reveal unique neoantigen profiles in Kenyan breast cancer patients

Neoantigen, breast cancer, exome-seq, RNA-seq, somatic mutations 14 15 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 Wagutu et al. Neoantigens profiling in Kenyan BC patients

16

The immune response against tumors relies on distinguishing between self and non-17 self, the basis of cancer immunotherapy. Neoantigens from somatic mutations are central to many 18 immunotherapeutic strategies and understanding their landscape in breast cancer is crucial for 19 targeted interventions. We aimed to profile neoantigens in Kenyan breast cancer patients using 20 genomic DNA and total RNA from paired tumor and adjacent non-cancerous tissue samples of 23 21 patients. 22

We sequenced the genome-wide exome (WES) and RNA, from which somatic mutations 23 were identified and their expression quantified, respectively. Neoantigen prediction focused on 24 human leukocyte antigens (HLA) crucial to cancer, HLA type I. HLA alleles were predicted from 25 WES data covering the adjacent non-cancerous tissue samples, identifying four alleles that were 26 present in at least 50% of the patients. Neoantigens were deemed potentially immunogenic if their 27 predicted median IC50 binding scores were ≤500nM and were expressed [transcripts per million 28 (TPM) >1] in tumor samples. 29

An average of 1465 neoantigens covering 10260 genes had ≤500nM median IC50 binding 30 score and >1 TPM in the 23 patients and their presence significantly correlated with the somatic 31 mutations (R2 =0.570, P=0.001). Assessing 58 genes reported in the catalog of somatic mutations in 32 cancer (COSMIC, v99) to be commonly mutated in breast cancer, 44 (76%) produced >2 neoantigens 33 among the 23 patients, with a mean of 10.5 ranging from 2 to 93. For the 44 genes, a total of 477 34 putative neoantigens were identified, predominantly derived from missense mutations (88%), indels 35 (6%), and frameshift mutations (6%). Notably, 78% of the putative breast cancer neoantigens were 36 patient-specific. HLA-C*06:01 allele was associated with the majority of neoantigens (194), 37 followed by HLA-A*30:01 (131), HLA-A*02:01 (103), and HLA-B*58:01 (49). Among the genes of 38 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 3 Wagutu et al. Neoantigens profiling in Kenyan BC patients interest that produced putative neoantigens were MUC17, TTN, MUC16, AKAP9, NEB, RP1L1, 39 CDH23, PCDHB10, BRCA2, TP53, TG, and RB1. 40

The unique neoantigen profiles in our patient group highlight the potential of 41 immunotherapy in personalized breast cancer treatment as well as potential biomarkers for prognosis. 42 The unique mutations producing these neoantigens, compared to other populations, provide an 43 opportunity for validation in a much larger sample cohort. 44 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 4 Wagutu et al. Neoantigens profiling in Kenyan BC patients

45 Breast cancer is among the most frequent causes of cancer-related mortality in women. Disease 46 heterogeneity and limited immunogenicity contribute to the lethality of breast cancer (Benvenuto et al., 47 2019). Immune evasion, an important hallmark of cancer, adds to the complexity of cancer burden 48 through induction of immunosuppression (Bates et al., 2018 ). Immune checkpoint blockade (CKB) 49 therapy has been developed to target and block immune regulatory molecules (PD -1/PD-L1 and 50 CTLA-4) and in the process reactivate T cell immunity (Touchaei &Vahidi, 2024). This approach has 51 been reported to improve clinic al responses and survival, especially in tumors with high mutational 52 burdens, such as lung cancer and melanoma (Shiravand et al., 2022). However, CKB therapy is not 53 universally successful among all patients and shows increased efficacy with higher mutational burden 54 tumors (Brahmer et al., 2015). Another immunotherapy approach that has been tested in clinical studies 55 is the targeting of tumor -associated antigens (TAAs) that are expressed in tumors at abnormally high 56 levels and rarely detectable in normal ti ssues (Valilou & Rezaei, 2019). One of the limitations of this 57 therapy approach is that many TAAs represent normal self -antigens and thus can be tolerated by T -58 cells, resulting in poor immune response (Benvenuto et al., 2019). This poses challenges for 59 applicability in breast cancer because it generally has a lower mutational burden. Thus, CKB and TAAs 60 immunotherapy have had limited success in breast cancer patients (Narang et al., 2019). 61 Tumor neoantigens are tumor -specific antigens derived from somatic m utations in expressed 62 genes and are presentable to the major histocompatibility complex (MHC) by both class I human 63 leukocyte antigen (HLA -I) molecules present on surface of cancer cell, as well as class II HLA 64 molecules present on professional antigen-presenting cells (Blass & Ott, 2021). This elicits anti-tumor 65 immune responses that have the potential of eliminating the tumor cells with minimal off-target effects 66 (Pan et al., 2018). Neoantigens are encoded in various mutational types, including single nuc leotide 67 substitution, insertion and deletions (INDELs), splice sites, stop codons gains and silent change, which 68 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 5 Wagutu et al. Neoantigens profiling in Kenyan BC patients can result in translational frameshifts or novel open reading frames (Benvenuto et al., 2019). As such, 69 these neoantigens offer an advantage ov er TAAs in that they are only expressed by cancer cells and 70 not by normal cells, which enables specific recognition by the immune system (Benvenuto et al., 2019). 71 Although some neoantigens are shared among patients, most of them are patient -specific and are not 72 subject to immune tolerance mechanisms (Yarchoan et al., 2017). The specificity of neoantigens could 73 provide an opportunity for future personalized therapy in a cancer with a low tumor mutational burden 74 and a high disease heterogeneity, such as breast cancer. Moreover, neoantigens can potentially be used 75 as biomarkers in cancer immunotherapy to assess or predict the response of a patient to treatment 76 (Benvenuto et al., 2019). 77 Despite advancements in next generation sequencing and high-performance computing that has 78 resulted in improved cancer immunotherapy research and neoantigen -based treatments, there remains 79 a scarcity of information regarding neoantigens in specific populations from sub -Saharan African 80 countries such as Kenya. This lack of data poses a significant challenge in tailoring immunotherapeutic 81 strategies for breast cancer patients in such regions that have a high cancer burden, especially when 82 compounded by germline ancestral factors and a distinct mutational spectrum that may influence tumor 83 biology and immune response. Thus, it is critical to profile the neoantigen burden in this population to 84 contribute to the global collection of breast cancer immunogenic antigens for future drug development. 85 To this end, we sought to profile neoantigens in Kenyan women diagnosed with breast cancer in silico 86 through analysis of the whole exome and RNA sequencing data from 23 patients. We characterized the 87 mutation burden for each patient using WES, identified gene expression patterns in tumor tissue, a nd 88 predicted the putative neoantigens incorporating these datasets. 89 90 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 6 Wagutu et al. Neoantigens profiling in Kenyan BC patients

91 Patients and samples 92 Tumor and adjacent normal tissue pairs were obtained from 23 breast cancer patients at the Aga 93 Khan Hospital, Nairobi, Kenya and AIC Kijabe Hospital, Kijabe, Kenya between 2019 and 2021. 94 Samples were collected through surgical excision, after which tissues were snap frozen in liquid 95 nitrogen and temporarily stored at Aga Khan Hospital. Frozen tissue samples were shipped to the 96 National Cancer Institute, Bethesda, MD, USA, for sequencing. Prior to tissue collection, all patients 97 provided written informed consent and the study was approved by Research and Ethics Committees at 98 Aga Khan University Hospital, Nairobi (Ref: 2018/REC -80) and AIC Kijabe Hospital (KH IERC -99 02718/2019). 100 Whole-exome sequencing (WES) and RNA-sequencing 101 Genomic DNA was extracted from the samples using the DNeasy Blood and Tissue Kit (Qiagen, 102 Hilden, Germany), following manufacturer’s instructions. Total RNA was extracted from the frozen 103 tissues using TRIzol reagent (Invitrogen). WES was performed by the company, Psomagen 104 (https://www.psomagen.com/). This service provider is Clinical Laboratory Improvement 105 Amendments-certified and College of American Pathologists (CAP)-accredited, achieving a sequence 106 depth of 250x for tumor tissues and 150x for adjacent non-cancerous tissues, as previously described 107 by us (Tang et al., 2023). Total RNA from the 23 sample pairs was processed by a NCI Leidos core 108 facility, where library preparation was performed using the TruSeq Poly A kit (Illumina, San Diego, 109 USA ). Samples were sequenced on a Novaseq system with 150 bp paired-end reads and a depth of 30 110 million reads. 111 Reads mapping and variant calling 112 For WES, raw reads were quality checked using FASTQC (Andrews, 2010) and results 113 summarized using MultiQC (Ewels et al., 2016). The reads were trimmed for low quality reads and 114 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 7 Wagutu et al. Neoantigens profiling in Kenyan BC patients adapter sequences using Trimmomatic (Bolger et al., 2014) and quality-checked again using FASTQC 115 and MultiQC. All samples passed the QC test after trimming and the reads were aligned using BWA -116 MEM (Li, 2013) to the hg38 human reference genome, where >95% of the reads aligned properly to 117 the genome. The aligned reads were deduplicated and read groups added to the deduplicated bam files 118 using Picard. This was followed by base quality recalibration in GATK (McKenna et al., 2010). 119 Somatic variant calling was performed using MuTect2 (McKenna et al., 2010) in paired tumor-normal 120 mode utilizing the panel of normal option that was derived from normal reads. Variants were 121 normalized using a variant tool set (vt; Tan et al., 2015), filtered using GATK and 122 functional/consequence-annotated using a variant effect predictor (VEP; McLaren et a l., 2016). 123 Annotated variants were converted to MAF files using vcf2maf (Kandoth et al., 2020) and concatenated 124 into a single file. The MAF files were imported into R package maftools (Mayakonda et al., 2018) for 125 further processing. 126 For RNA-seq, a quality check was performed using FASTQC and MultiQC after which the 127 reads were trimmed and quality checked again. All samples passed the quality check and the reads 128 were pseudo -aligned to the hg38 reference genome using Kallisto aligner (Bray et al., 2016) with 129 default settings to obtain count matrix. Alignment statistics showed that over >50% reads mapped 130 uniquely to the genome. The raw counts were normalized into estimated Transcripts Per Million (TPM), 131 and scaled using the average transcript length over samples and t he library size by tximport (Soneson 132 et al., 2016). 133 Variant expression annotation 134 VCF files containing the variants were annotated for expression using the vcf -expression-135 annotator (https://github.com/griffithlab/VAtools) with default setting except for c hoosing the use of 136 gene names instead of transcripts and thereby ignoring the Ensembl id version. The tool takes the 137 output of Kallisto and adds the data contained in the file to the VEP annotated VCF’s INFO column. 138 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 8 Wagutu et al. Neoantigens profiling in Kenyan BC patients Each of the variant annotated gets its expression value (TPM) added to the annotation information and 139 this is used to determine the level of variant expression during neoantigen filtering. 140 Neoantigen prediction 141 Human leukocyte antigen (HLA) class I alleles (HLA a, b and c) were predicted from ea ch 142 patient’s normal sample exome -seq data using HLA -HD v.1.2.1 (Kawaguchi et al., 2017). Here, the 143 putative HLA reads are aligned to an imputed library of full -length HLA alleles. Neoantigens were 144 then predicted using pVACseq (Hundal et al, 2016) with MHCflurry, MHCnuggetsI, SMM, and 145 SMMPMBEC algorithms and keeping the default parameters, except for turning off the VAF and 146 coverage filters. Here, the neoepitopes that c ould bind to the patient -specific HLA alleles were 147 predicted from the Immune Epitope Database (IEDB; Vita et al., 2019). This involved matching patient 148 HLA type to the existing IEDB list keeping all amino acids with lengths for 9, 10 and 11 -mers. 149 Predicted epitopes were filtered to retain only those with high affinity (IC50 ≤ 500nM) and were 150 expressed (transcripts per million, TPM>1) in tumor samples. The bioinformatic analysis workflow is 151 outlined in Figure 1. 152 Sample summary statistics and the pairwise tests for differences among mutations and neoan-153 tigens abundance among the BC subtypes using Wilcoxon test and visualization of the results were 154 performed in R software (R Core Team, 2023). 155

156 Patients and sample characteristics 157 The demographic and clinical characteristics of the 23 breast cancer patients are summarized 158 in supplementary Table S1. We grouped the tumors into 3 subtypes based on e xpression of either the 159 hormone receptors (HR) or human epidermal growth factor receptor 2 (HER2) (Narang et al., 2019): 160 those that were HER2+ regardless of the HR status, those that were negative for all hormone receptors 161 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 9 Wagutu et al. Neoantigens profiling in Kenyan BC patients (triple negative breast cancer; T NBC) and those that were HR+ and HER2 -. Majority of the samples 162 were HR+/HER2- constituting 52.2%, followed by HER2+ at 34.8% and TNBC at 13.0%. Most of the 163 patients had invasive carcinoma (invasive ductal carcinoma, 78.26% and invasive carcinoma; 4.35%). 164 For tumor grade, 65.22% of the patients had grade 3 tumors (65.22%), while the rest had grade 2 tumors 165 (34.78%). Clinically, 39.13% of the patients were in stage II, 30.44% in stage III, and 8.7% in stage I 166 (Table S1). 167 Mutation profiles for the 23 patients 168 Across all genes, the average number of detected mutations in the 23 patients was 2809 169 mutations. Considering the different subtypes, TNBC had the highest average number of mutations at 170 3202, followed by HR+/HER2- at 2757, and HER2+ at 2740 mutations (Figure S1). From the catalog 171 of somatic mutations in cancer (COSMIC, v99), we identified 73 genes reported to be mutated in breast 172 cancer and among those, 62 (84.9%) had at least one mutation in our samples. The mutation frequency 173 among the 62 genes ranged from 1 to 55 mutations per individual. The majority of the mutations were 174 of the missense type, most of which were substitutions of C>T (Figure 2). The top 10 mutated genes 175 among the 62 are shown in Figure 3. Four genes (MUC16, MUC17, TTN, RP1L1) were altered in more 176 than 95% of the patients (Figure 3). Moreover, mutations in genes TP53-ERBB3, PTEN-CFAP46 were 177 found to significantly co-occur, while BRCA1-MUC17 mutations were significantly mutually exclusive 178 (P<0.05) (Figure 4). Furthermore, the majority of the single nucleotide mutations were substitution of 179 C to T, whereas T to A substitutions were most uncommon. Transitions occurred more frequently than 180 transversion in these substitutions (Figure 5). 181 Neoantigen burden 182 In an analysis that included all the genes (10260), an average of 1465 neoantigens had a 183 ≤500nM median IC50 binding score and >1 TPM expression level in any of the 23 patients and their 184 presence significantly correlated with the somatic mutations ( R2=0.570, P=0.001) (Figure 6). Out of 185 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 10 Wagutu et al. Neoantigens profiling in Kenyan BC patients the 62 COSMIC genes that were mutated in the tumor tissue, 58 genes produced at least one neoantigen. 186 After filtering for genes that produced at least two neoantigens, 44 genes had a mean of 10.5 187 neoantigens ranging from 2 to 93. A total of 477 putative neoantigens were identified in these 44 genes 188 across the 23 patients (Figure 7) predominantly derived from missense mutations (88%), indels (6%) 189 and frameshift mutations (6%) (Figure 8). Most of the neoantigens were produced in the TNBC subtype 190 with an average of 25 neoantigens, followed b y HR+/HER2- at 20 neoantigens and HER2+ with an 191 average of 19 neoantigens (Figure S1). Notably, 78% of the putative breast cancer neoantigens were 192 patient-specific (Table S2). HLA -C*06:01 allele was associated with majority of neoantigens (194), 193 followed by HLA-A*30:01 (131), HLA-A*02:01 (103), and HLA-B*58:01 (49). Among the genes of 194 interest that produced putative neoantigens include MUC17, TTN, MUC16, AKAP9, NEB, RP1L1, 195 CDH23, PCDHB10, BRCA2, TP53, TG, RB1 among others (Figure 7, Table S3). 196 197 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 11 Wagutu et al. Neoantigens profiling in Kenyan BC patients

198 We analyzed the mutational burden and predicted the neoantigen repertoire in 23 Kenyan breast 199 cancer patients using WES and RNA sequencing data. Among the different breast cancer subtypes, we 200 found that the TNBC molecular subtype had the highest mutational and neoantigen burden although 201 there was no significant difference among the subtypes (Figure S1, Table S4). This is consistent with 202 other studies (Narang et a., 2019). TNBC origin is not well understood although it is reported to be 203 heterogeneous in natu re relying on different signaling pathways such as JAK/STAT, 204 PI3K/AKT/mTOR or NOTCH, cell cycle regulators ( TP53) and genome integrity genes ( BRCA1/2) 205 (Benvenuto et al., 2019). This makes it a disease that is difficult to manage because we do not have a 206 clear understanding of the molecular mechanisms driving it. Yet, the high mutational and neoantigens 207 burden combined with the patient specificity may provide an untapped opportunity to design and 208 optimize personalized immunotherapy for this subtype. 209 In contrast to most populations where TP53, PIK3CA and GATA3 are the most mutated genes 210 (Pan et al., 2020; Pipek et al., 2023; Tang et al., 2023), in our study population, three genes MUC16, 211 MUC17 and TTN were highly mutated in over 50% of the samples and produc ed the highest number 212 of neoantigens. MUC16 has been reported to take part in breast cancer progression and metastasis when 213 overexpressed due to its influence on cell cycle and survival through the JAK2/STAT3 pathway 214 (Lakshmanan et al., 2012). It has been reported as one of the highly mutated genes in breast cancer 215 (Wang & Guda, 2016). MUC16 has also been described as a marker for disease progression, recurrence, 216 and chemotherapy response (Felder et al., 2014). A high mutation frequency for MUC17 and TTN have 217 recently been reported as an unexpected finding in a study of early onset breast cancer (EOBC) in 218 Taiwanese women (Midha et al., 2020). MUC17 may influence chemoresistance and has recently been 219 reported as a driver gene in adult gliomas (Al Amri et al., 2020; Machado & Ferrer, 2023). For TTN, 220 Oh et al. (2020) found that mutations in TTN correlate with tumor mutational burden and high 221 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 12 Wagutu et al. Neoantigens profiling in Kenyan BC patients microsatellite instability, which is associated with poor breast cancer prognosis. Thus, the role of 222 MUC17 and TTN should further be investigated on how mutations in them may relate to early onset of 223 breast cancer in Kenyan patients (Tang et al., 2023). 224 We found that TP53 gene mutations significantly co -occurred with ERBB3 mutations and so 225 did mutations in PTEN and CFAP46, whereas BRCA1 and MUC17 mutations never co-occurred. TP53 226 mutations are associated with tumor aggression and are found in about half of HER2-amplified tumors 227 (Marvalim et al., 2023). The TP53 mutations have been implicated in poor prognosis of HER2+ 228 subtypes compared to other subtypes (Dumay et al., 2013). PTEN is a tumor suppressor gene, whose 229 mutation has been associated with initiation, progression, and metastasis of breast cancer (Chen et al., 230 2022). On the other hand, although CFAP46 role in breast cancer is not yet clear, gene fusion involving 231 various other genes such as VTI1A (reported to cause the initiation of glioma and other cancers) has 232 been reported to play a role in breast cancer (Tsuge et al., 2019). 233 Breast tumors with either germline or somatic BRCA1 mutations show no difference in th eir 234 cancer biology, but inherited mutations in this gene confers a very high lifetime risk of developing 235 breast cancer (Milne & Antoniou, 2011; den Brok et al., 2017; Bodily et al., 2020). This could be the 236 reason such mutations do not necessarily need to co-occur with other gene mutations to initiate or 237 promote breast cancer progression. In our study, BRCA1 was not among the highly mutated genes 238 considering all mutations but was among the genes with high number of missense mutations (Figure 239 4). In contrast, MUC17 mutations were among the most prevalent. Given the role of MUC17 mutations 240 in chemoresistance and in early onset breast cancer (Al Amri et al., 2020; Machado & Ferrer, 2023), 241 its high prevalence and exclusive occurrence in the Kenyan samples that are prone to early onset of 242 breast cancer should be investigated further. 243 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 13 Wagutu et al. Neoantigens profiling in Kenyan BC patients Similar to most studies on neoantigen prediction in breast cancer, we have found that 244 neoantigens burden is positively correlated with tumor mutational burden and that neoantigens we re 245 patient-specific (Narang et al., 2019; Animesh et al., 2022). Although most of the top 10 mutated genes 246 (80%) were also the top 10 in the number of neoantigens generated, genes like TP53 and PIK3CA that 247 are reported to be highly mutated in most patient cohorts were not among the top 10 mutated genes in 248 this study, but generated among the highest number of neoantigens (Figure 6; Figure 7). ARID1A gene, 249 which showed unique mutational profile in Kenyan population using exome data compared to African 250 American and Asian population (Tang et al., 2023), was not among the highly mutated, but produced 251 neoantigens. We found that most neoantigens were derived predominantly from missense mutations 252 (88%), compared to indels and frameshift mutations (12%). This is cons istent with other studies 253 although the majority do not predict neoantigens from indels and frameshift mutations (Morisaki et al., 254 2021). Similar to other studies, the TNBC subtype had more neoantigens, compared to HR+/HER2 - 255 and HER2+ subtypes (Narang et al., 2019; Morisaki et al., 2021). 256 In our small sample cohort, we have been able to identify putative neoantigens that show 257 patient-specificity and thus are important in tailored treatment. Interestingly, the mutations and 258 neoantigens in this population are predominantly derived from a unique set of genes (MUC16, MUC17, 259 TNT) compared to other populations, which provide an opportunity for validation in a much larger 260 sample cohort. We predicted neoantigens based on binding affinity to HLA class I only as it is the most 261 important class of antigen binding proteins in cancer immunity. However, HLA class II -based 262 neoantigens may also have a role in tumor immune response (Alspach et al., 2019). Moreover, we did 263 not investigate the expression of the predicted neoanti gens on tumor cells alongside the MHC class I 264 molecules and their ability to activate T cells. This being a discovery study, validation of the findings 265 need to be done in a larger cohort while addressing the highlighted limitations of this study. 266 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 14 Wagutu et al. Neoantigens profiling in Kenyan BC patients Taken together, our findings corroborate the neoantigen profile in breast cancer, highlighting 267 the patient specificity in Kenyan population breast cancer mutational and neoantigens signatures. We 268 also describe putative neoantigens that could be used as markers for breast cancer diagnosis, treatment 269 monitoring, and development of novel immunotherapy. 270 Acknowledgment 271 We would like to thank the patients for their consent to provide samples and Aga Khan University 272 Hospital (Nairobi) and AIC Kijabe Hospital (Kijabe) for granting access to patient samples. 273 Funding 274 This work was funded by the National Research Fund – Kenya that supported sample collection, and 275 by the Center for Cancer Research, National Cancer Institute, USA, that supported the sequencing 276 work. 277 Data accessibility 278 WES data is accessible at SRA database Accession number: PRJNA913947, while RNA -seq data is 279 accessible at the GEO database under Accession number: GSE225846. All other datasets for this study 280 are included in the article’s Supplementary Material. 281 Authors’ contribution 282 FM conceived the idea and designed the research project, collected the samples and assembled 283 experiment materials. GW, JG, KM and MM performed the data analysis. FM and GW wrote the 284 manuscript. ARH, SS, SA helped with drafting and reviewing the manuscript. All authors contributed 285 to the revision and final editing of the manuscript prior to submission. 286 287 288 289 290 291 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 15 Wagutu et al. Neoantigens profiling in Kenyan BC patients

292 Andrews S. (2010). FastQC: a quality control tool for high throughput sequence data. Available 293 online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc 294 Al Amri, W. S., Allinson, L. M., Baxter, D. E., Bell, S. M., Hanby, A. M., Jones, S. J., Shaaban, A. 295 M., Stead, L. F., Verghese, E. T., & Hughes, T. A. (2020). Genomic and Expression Analyses Define 296 MUC17 and PCNX1 as Predictors of Chemotherapy Response in Breast Cancer. Molecular Cancer 297 Therapeutics, 19(3), 945–955. https://doi.org/10.1158/1535-7163.MCT-19-0940 298 Alspach, E., Lussier, D. M., Miceli, A. P., Kizhvatov, I., DuPage, M., Luoma, A. M., Meng, W., 299 Lichti, C. F., Esaulova, E., Vomund, A. N., Runci, D., Ward, J. P., Gubin, M. M., Medrano, R. F. V., 300 Arthur, C. D., White, J. M., Sheehan, K. C. F., Chen, A., Wucherpfennig, K. W., … Schreiber, R. D. 301 (2019). MHC-II neoantigens shape tumour immunity and response to immunotherapy. Nature, 302 574(7780), 696–701. https://doi.org/10.1038/s41586-019-1671-8 303 Animesh, S., Ren, X., An, O., Chen, K., Lee, S. C., Yang, H., & Fullwood, M. J. (2022). Exploring 304 the Neoantigen burden in Breast Carcinoma Patients. https://doi.org/10.1101/2022.03.03.482669 305 Bates, J. P., Derakhshandeh, R., Jones, L., & Webb, T. J. (2018). Mechanisms of immune evasion in 306 breast cancer. BMC cancer, 18(1), 556. https://doi.org/10.1186/s12885-018-4441-3 307 Benvenuto, M., Focaccetti, C., Izzi, V., Masuelli, L., Modesti, A., & Bei, R. (2021). Tumor antigens 308 heterogeneity and immune response-targeting neoantigens in breast cancer. Seminars in cancer 309 biology, 72, 65–75. https://doi.org/10.1016/j.semcancer.2019.10.023 310 Benvenuto, M., Focaccetti, C., Izzi, V., Masuelli, L., Modesti, A., & Bei, R. (2021). Tumor antigens 311 heterogeneity and immune response-targeting neoantigens in breast cancer. Seminars in Cancer 312 Biology, 72, 65–75. https://doi.org/10.1016/j.semcancer.2019.10.023 313 Blass, E., & Ott, P. A. (2021). Advances in the development of personalized neoantigen-based 314 therapeutic cancer vaccines. Nature Reviews Clinical Oncology, 18(4), 215–229. 315 https://doi.org/10.1038/s41571-020-00460-2 316 Bodily W.R., Shirts B.H., Walsh T., Gulsuner S., King M.-C., Parker A., Roosan M., Piccolo S.R. 317 Effects of Germline and Somatic Events in Candidate BRCA-like Genes on Breast-Tumor 318 Signatures. PLoS ONE. 2020;15:e0239197. doi: 10.1371/journal.pone.0239197 319 Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: A flexible trimmer for Illumina 320 sequence data. Bioinformatics, 30(15), 2114–2120. https://doi.org/10.1093/bioinformatics/btu170 321 Brahmer, J., Reckamp, K. L., Baas, P., Crinò, L., Eberhardt, W. E. E., Poddubskaya, E., Antonia, S., 322 Pluzanski, A., Vokes, E. E., Holgado, E., Waterhouse, D., Ready, N., Gainor, J., Arén Frontera, O., 323 Havel, L., Steins, M., Garassino, M. C., Aerts, J. G., Domine, M., … Spigel, D. R. (2015). 324 Nivolumab versus Docetaxel in Advanced Squamous-Cell Non–Small-Cell Lung Cancer. New 325 England Journal of Medicine, 373(2), 123–135. https://doi.org/10.1056/NEJMoa1504627 326 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 16 Wagutu et al. Neoantigens profiling in Kenyan BC patients Bray, N. L., Pimentel, H., Melsted, P., & Pachter, L. (2016). Near-optimal probabilistic RNA-seq 327 quantification. Nature Biotechnology, 34(5), 525–527. https://doi.org/10.1038/nbt.3519 328 Chen, J., Sun, J., Wang, Q., Du, Y., Cheng, J., Yi, J., Xie, B., Jin, S., Chen, G., Wang, L., Wang, X., 329 & Wei, H. (2022). Systemic Deficiency of PTEN Accelerates Breast Cancer Growth and Metastasis. 330 Frontiers in Oncology, 12, 825484. https://doi.org/10.3389/fonc.2022.825484 331 den Brok W.D., Schrader K.A., Sun S., Tinker A.V., Zhao E.Y., Aparicio S., Gelmon K.A. 332 Homologous Recombination Deficiency in Breast Cancer: A Clinical Review. JCO Precis. Oncol. 333 2017;1:1–13. doi: 10.1200/PO.16.00031. 334 Dumay, A., Feugeas, J., Wittmer, E., Lehmann-Che, J., Bertheau, P., Espié, M., Plassa, L., Cottu, P., 335 Marty, M., André, F., Sotiriou, C., Pusztai, L., & De Thé, H. (2013). Distinct tumor protein p53 336 mutants in breast cancer subgroups. International Journal of Cancer, 132(5), 1227–1231. 337 https://doi.org/10.1002/ijc.27767 338 Ewels, P., Magnusson, M., Lundin, S., & Käller, M. (2016). MultiQC: Summarize analysis results for 339 multiple tools and samples in a single report. Bioinformatics, 32(19), 3047–3048. 340 https://doi.org/10.1093/bioinformatics/btw354 341 Felder, M., Kapur, A., Gonzalez-Bosquet, J., Horibata, S., Heintz, J., Albrecht, R., Fass, L., Kaur, J., 342 Hu, K., Shojaei, H., Whelan, R. J., & Patankar, M. S. (2014). MUC16 (CA125): Tumor biomarker to 343 cancer therapy, a work in progress. Molecular Cancer, 13(1), 129. https://doi.org/10.1186/1476-344 4598-13-129 345 Hundal, J., Carreno, B. M., Petti, A. A., Linette, G. P., Griffith, O. L., Mardis, E. R., & Griffith, M. 346 (2016). pVAC-Seq: A genome-guided in silico approach to identifying tumor neoantigens. Genome 347 Medicine, 8(1), 11. https://doi.org/10.1186/s13073-016-0264-5 348 Kawaguchi, S., Higasa, K., Shimizu, M., Yamada, R., & Matsuda, F. (2017). HLA-HD: An accurate 349 HLA typing algorithm for next-generation sequencing data. Human Mutation, 38(7), 788–797. 350 https://doi.org/10.1002/humu.23230 351 Kandoth, C., Gao, J., Qwangmsk, Mattioni, M., Struck, A., Boursin, Y., Penson, A., & Chavan, S. 352 (2018). mskcc/vcf2maf: Vcf2maf v1.6.16 (v1.6.16) [Computer software]. 353 https://doi.org/10.5281/ZENODO.593251 354 Lakshmanan, I., Ponnusamy, M. P., Das, S., Chakraborty, S., Haridas, D., Mukhopadhyay, P., Lele, 355 S. M., & Batra, S. K. (2012). MUC16 induced rapid G2/M transition via interactions with JAK2 for 356 increased proliferation and anti-apoptosis in breast cancer cells. Oncogene, 31(7), 805–817. 357 https://doi.org/10.1038/onc.2011.297 358 Li, H. (2013). Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. 359 https://doi.org/10.48550/ARXIV.1303.3997 360 Machado, G. C., & Ferrer, V. P. (2023). MUC17 mutations and methylation are associated with poor 361 prognosis in adult-type diffuse glioma patients. Journal of the Neurological Sciences, 452, 120762. 362 https://doi.org/10.1016/j.jns.2023.120762 363 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 17 Wagutu et al. Neoantigens profiling in Kenyan BC patients Marvalim, C., Datta, A., & Lee, S. C. (2023). Role of p53 in breast cancer progression: An insight 364 into p53 targeted therapy. Theranostics, 13(4), 1421–1442. https://doi.org/10.7150/thno.81847 365 Mayakonda, A., Lin, D.-C., Assenov, Y., Plass, C., & Koeffler, H. P. (2018). Maftools: Efficient and 366 comprehensive analysis of somatic variants in cancer. Genome Research, 28(11), 1747–1756. 367 https://doi.org/10.1101/gr.239244.118 368 McKenna, A., Hanna, M., Banks, E., Sivachenko, A., Cibulskis, K., Kernytsky, A., Garimella, K., 369 Altshuler, D., Gabriel, S., Daly, M., & DePristo, M. A. (2010). The Genome Analysis Toolkit: A 370 MapReduce framework for analyzing next-generation DNA sequencing data. Genome Research, 371 20(9), 1297–1303. https://doi.org/10.1101/gr.107524.110 372 McLaren, W., Gil, L., Hunt, S. E., Riat, H. S., Ritchie, G. R. S., Thormann, A., Flicek, P., & 373 Cunningham, F. (2016). The Ensembl Variant Effect Predictor. Genome Biology, 17(1), 122. 374 https://doi.org/10.1186/s13059-016-0974-4 375 Midha, M. K., Huang, Y.-F., Yang, H.-H., Fan, T.-C., Chang, N.-C., Chen, T.-H., Wang, Y.-T., Kuo, 376 W.-H., Chang, K.-J., Shen, C.-Y., Yu, A. L., Chiu, K.-P., & Chen, C.-J. (2020). Comprehensive 377 Cohort Analysis of Mutational Spectrum in Early Onset Breast Cancer Patients. Cancers, 12(8), 378 2089. https://doi.org/10.3390/cancers12082089 379 Milne, R. L. & Antoniou, A. C. (2011). Genetic modifiers of cancer risk for BRCA1 and BRCA2 380 mutation carriers. Annals of Oncology, 22 (1): i11-i17. 381 Morisaki, T., Kubo, M., Umebayashi, M., Yew, P. Y., Yoshimura, S., Park, J.-H., Kiyotani, K., Kai, 382 M., Yamada, M., Oda, Y., Nakamura, Y., Morisaki, T., & Nakamura, M. (2021). Neoantigens elicit T 383 cell responses in breast cancer. Scientific Reports, 11(1), 13590. https://doi.org/10.1038/s41598-021-384 91358-1 385 Narang, P., Chen, M., Sharma, A. A., Anderson, K. S., & Wilson, M. A. (2019). The neoepitope 386 landscape of breast cancer: Implications for immunotherapy. BMC Cancer, 19(1), 200. 387 https://doi.org/10.1186/s12885-019-5402-1 388 Oh, J.-H., Jang, S. J., Kim, J., Sohn, I., Lee, J.-Y., Cho, E. J., Chun, S.-M., & Sung, C. O. (2020). 389 Spontaneous mutations in the single TTN gene represent high tumor mutation burden. Npj Genomic 390 Medicine, 5(1), 33. https://doi.org/10.1038/s41525-019-0107-6 391 Pan, J.-W., Zabidi, M. M. A., Ng, P.-S., Meng, M.-Y., Hasan, S. N., Sandey, B., Sammut, S.-J., Yip, 392 C.-H., Rajadurai, P., Rueda, O. M., Caldas, C., Chin, S.-F., & Teo, S.-H. (2020). The molecular 393 landscape of Asian breast cancers reveals clinically relevant population-specific differences. Nature 394 Communications, 11(1), 6433. https://doi.org/10.1038/s41467-020-20173-5 395 Pan, R.-Y., Chung, W.-H., Chu, M.-T., Chen, S.-J., Chen, H.-C., Zheng, L., & Hung, S.-I. (2018). 396 Recent Development and Clinical Application of Cancer Vaccine: Targeting Neoantigens. Journal of 397 Immunology Research, 2018, 1–9. https://doi.org/10.1155/2018/4325874 398 Pipek, O., Alpár, D., Rusz, O., Bödör, C., Udvarnoki, Z., Medgyes-Horváth, A., Csabai, I., Szállási, 399 Z., Madaras, L., Kahán, Z., Cserni, G., Kővári, B., Kulka, J., & Tőkés, A. M. (2023). Genomic 400 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 18 Wagutu et al. Neoantigens profiling in Kenyan BC patients Landscape of Normal and Breast Cancer Tissues in a Hungarian Pilot Cohort. International Journal 401 of Molecular Sciences, 24(10), 8553. https://doi.org/10.3390/ijms24108553 402 R Core Team (2023). R: A language and environment for statistical computing. R Foundation for 403 Statistical Computing, Vienna, Austria. URL https://www.R-project.org/ 404 Shiravand, Y., Khodadadi, F., Kashani, S. M. A., Hosseini-Fard, S. R., Hosseini, S., Sadeghirad, H., 405 Ladwa, R., O’Byrne, K., & Kulasinghe, A. (2022). Immune Checkpoint Inhibitors in Cancer 406 Therapy. Current Oncology, 29(5), 3044–3060. https://doi.org/10.3390/curroncol29050247 407 Soneson, C., Love, M. I., & Robinson, M. D. (2016). Differential analyses for RNA-seq: Transcript-408 level estimates improve gene-level inferences. F1000Research, 4, 1521. 409 https://doi.org/10.12688/f1000research.7563.2 410 Tan, A., Abecasis, G. R., & Kang, H. M. (2015). Unified representation of genetic variants. 411 Bioinformatics, 31(13), 2202–2204. https://doi.org/10.1093/bioinformatics/btv112 412 Tang, W., Zhang, F., Byun, J. S., Dorsey, T. H., Yfantis, H. G., Ajao, A., Liu, H., Pichardo, M. S., 413 Pichardo, C. M., Harris, A. R., Yang, X. R., Figueroa, J. D., Sayed, S., Makokha, F. W., & Ambs, S. 414 (2023). Population-specific Mutation Patterns in Breast Tumors from African American, European 415 American, and Kenyan Patients. Cancer Research Communications, 3(11), 2244–2255. 416 https://doi.org/10.1158/2767-9764.CRC-23-0165 417 Touchaei, Z. A., & Vahidi, S. (2024). MicroRNAs as regulators of immune checkpoints in cancer 418 immunotherapy: Targeting PD-1/PD-L1 and CTLA-4 pathways. Cancer Cell International, 24(1), 419 102. https://doi.org/10.1186/s12935-024-03293-6 420 Tsuge, S., Saberi, B., Cheng, Y., Wang, Z., Kim, A., Luu, H., Abraham, J. M., Ybanez, M. D., 421 Hamilton, J. P., Selaru, F. M., Villacorta-Martin, C., Schlesinger, F., Philosophe, B., Cameron, A. 422 M., Zhu, Q., Anders, R., Gurakar, A., & Meltzer, S. J. (2019). Detection of Novel Fusion Transcript 423 VTI1A-CFAP46 in Hepatocellular Carcinoma. Gastrointestinal Tumors, 6(1–2), 11–27. 424 https://doi.org/10.1159/000496795 425 Valilou, S. F., & Rezaei, N. (2019). Tumor Antigens. In Vaccines for Cancer Immunotherapy (pp. 426 61–74). Elsevier. https://doi.org/10.1016/B978-0-12-814039-0.00004-7 427 Vita, R., Mahajan, S., Overton, J. A., Dhanda, S. K., Martini, S., Cantrell, J. R., Wheeler, D. K., 428 Sette, A., & Peters, B. (2019). The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids 429 Research, 47(D1), D339–D343. https://doi.org/10.1093/nar/gky1006 430 Wang, X., & Guda, C. (2016). Integrative exploration of genomic profiles for triple negative breast 431 cancer identifies potential drug targets. Medicine, 95(30), e4321. 432 https://doi.org/10.1097/MD.0000000000004321 433 Yarchoan, M., Johnson, B. A., 3rd, Lutz, E. R., Laheru, D. A., & Jaffee, E. M. (2017). Targeting 434 neoantigens to augment antitumour immunity. Nature reviews. Cancer, 17(4), 209–222. 435 https://doi.org/10.1038/nrc.2016.154 436 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 19 Wagutu et al. Neoantigens profiling in Kenyan BC patients Figures 437 438 439 Figure 1: Workflow for neoantigen prediction from WES and RNA sequencing data. Fastq files 440 were quality checked, trimmed and aligned to the hg38 genome. Variant calling was performed 441 following GATK best practice, while gene expression was quantified using Kallisto. Variants were 442 annotated and expression data added, after which neoantigen prediction was performed in PVACseq 443 pipeline 444 445 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 20 Wagutu et al. Neoantigens profiling in Kenyan BC patients 446 Figure 2: Mutational profiles in 23 patients for 73 genes reported to be mutated in breast cancer. A) 447 variant classes abundance in the total mutations, B) variant types that include single nucleotide 448 polymorphism (SNP), insertions (INS) and deletions (DEL), C) proportion of different single 449 nucleotide variant (SNV), D) distribution of variants per sample with colors representing the different 450 variant classes denoted in A, E) summary of the variant classes distribution and numbers in all 451 samples, F) Top 10 mutated genes, with colors representing different variant classes and the 452 percentages indicating the proportion of samples in which the genes mutations are present. 453 454 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 21 Wagutu et al. Neoantigens profiling in Kenyan BC patients 455 Figure 3: Top 10 genes mutated in >50% of the samples. Each color corresponds to a variant class 456 listed at the bottom of the figure apart from gray, which indicates absence of mutation. 457 458 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 22 Wagutu et al. Neoantigens profiling in Kenyan BC patients 459 Figure 4: Probability of mutations in any two genes co-occurrence or being mutually exclusive in the 460 breast cancer genes for the 23 Kenyan patients. The numbers in parenthesis alongside each gene 461 represents the number of missense mutations for that gene in the samples. 462 463 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 23 Wagutu et al. Neoantigens profiling in Kenyan BC patients 464 Figure 5: A) Percentage of various substitution types in all samples, B) percentage of transversions 465 (interchange of purines for pyrimidine) and transition (interchange of either purines or pyrimidines) 466 for all samples, C) percentage of the substitutions in each of the samples with colors denoting the 467 various types in A. 468 469 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 24 Wagutu et al. Neoantigens profiling in Kenyan BC patients 470 Figure 6: Correlation between tumor mutational burden and neoantigen burden for all the genes in 471 the 23 patients. The neoantigens are filtered for high affinity (IC50 ≤ 500nM) and expression 472 (transcripts per million, TPM>1) in tumor samples. 473 474 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 25 Wagutu et al. Neoantigens profiling in Kenyan BC patients 475 Figure 7: Frequency of neoantigens derived from the COSMIC genes that were mutated in the tumor 476 tissue and produced >1 neoantigens for the 23 patients. 477 478 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 26 Wagutu et al. Neoantigens profiling in Kenyan BC patients 479 Figure 8: Summary of mutation types that produced putative neoantigens for the COSMIC genes 480 that were mutated in the tumor tissue in the 23 Kenyan patients. 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint 27 Wagutu et al. Neoantigens profiling in Kenyan BC patients Supplementary Materials 499 500 501 Figure S1: Statistical pairwise test (Wilcoxon’s test) for differences in mutational burden (A) and 502 neoantigens counts (B) for the 23 samples. 503 504 Table S1: Sample characteristics of the 23 Kenyan patients used in this study. 505 506 Table S2: Putative neoantigens for each of the 23 Kenyan patients. Cells in red indicate that the 507 neoantigen is shared by at least 2 patients. 508 509 Table S3: Summary of the roles of the top ten genes that generated a high number of neoantigens. 510 511 Table S4: Summary of total mutations and proportion of mutation types, total neoantigens, filtered 512 total neoantigens (filtered for high affinity (IC50 ≤ 500nM) and expression [transcripts per million, 513 TPM>1] in tumor samples) and filtered putative neoantigens from COSMIC 44 genes mutated in 514 tumor tissue, and proportion of mutation types that generated them per breast cancer subtype for the 515 23 samples 516 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 20, 2024. ; https://doi.org/10.1101/2024.06.18.24309133doi: medRxiv preprint