CADRES: An Analytical Pipeline for Precise Identification of Differential RNA Editing Events Across Varied Biological Conditions

CADRES: An Analytical Pipeline for Precise Identification of Differential RNA Editing Events Across Varied Biological Conditions | 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 CADRES: An Analytical Pipeline for Precise Identification of Differential RNA Editing Events Across Varied Biological Conditions Jun Sun, Chi Zhang, Xiuling Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4855669/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 04 Jun, 2025 Read the published version in Scientific Reports → Version 1 posted 11 You are reading this latest preprint version Abstract RNA editing is a post-transcriptional modification critical for gene regulation and adaptation. Detecting these modifications, especially D ifferential V ariants on R NA (DVRs), presents significant challenges due to the subtlety of RNA editing sites and the complexity of RNA sequences. To improve the detection and analysis of RNA editing sites, we developed the Ca librated D ifferential R NA E diting S canner (CADRES), an analytical pipeline that combines sophisticated DNA/RNA variant calling with detailed statistical analysis. This study validates CADRES through rigorous in silico and experimental dataset using inducible cell models of the APOBEC3B (A3B) deaminase. CADRES demonstrates improved specificity and accuracy over existing methodologies, effectively identifying A3B-mediated C > U edits and distinguishing these from related sequencing artifacts and DNA polymorphisms. The adaptability of CADRES is highlighted by its consistent performance across varied experimental conditions and different numbers of replicates. Our findings illustrate that CADRES provides a reliable tool for the precise identification of RNA editing sites, contributing valuable insights into RNA editing dynamics. This capability is crucial for advancing our understanding of the molecular mechanisms underlying gene expression regulation and the potential development of therapeutic strategies. Biological sciences/Computational biology and bioinformatics Biological sciences/Molecular biology/Rna metabolism/Rna editing Biological sciences/Molecular biology/Rna metabolism/Rna modification Biological sciences/Biotechnology/Sequencing/Dna sequencing Biological sciences/Biotechnology/Sequencing/Next generation sequencing Biological sciences/Biotechnology/Sequencing/Rna sequencing RNA editing Next-generation sequencing Differential Variants on RNA (DVRs) Bioinformatics Variant calling APOBEC deaminase Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction RNA editing is a crucial post-transcriptional modification that enables organisms to dynamically alter transcript sequences without modifying the underlying genomic DNA 1 . This process is fundamental in enhancing proteomic diversity and exerting regulatory effects on gene expression across higher eukaryotes 2 . At the core of this mechanism are enzymatic activities such as the conversion of adenine to inosine (A > I) and cytidine to uracil (C > U), which are recognized during translation as guanosine and uridine, respectively. These modifications can significantly alter the amino acid sequences encoded by the genomic DNA, potentially impacting protein function and cellular behaviour 3 , 4 . Beyond these well-known types, other RNA editing mechanisms, including those involving methylation and pseudouridylation, also play critical roles in cellular response to environmental changes and disease pathogenesis 5 . The expanding understanding of RNA editing, propelled by advancements in sequencing technologies 6 , underscores its importance in biological regulation and disease development. Among the most significant contributors to RNA editing are the ADAR ( A denosine D eaminases A cting on R NA) and APOBEC (Apolipoprotein B mRNA Editing Enzyme, Catalytic Polypeptide-like) families of deaminases 3 , 4 . ADAR enzymes, comprising ADAR1, ADAR2, and ADAR3, are essential for A > I editing and play a critical role in normal cellular function. They modulate the coding potential and expression of neurotransmitter receptors and ion channels, crucial for maintaining synaptic plasticity and proper neuronal signalling 7 . Beyond their role in the brain, ADAR enzymes are involved in the immune system, where they edit RNA transcripts encoding interferon-inducible proteins, thus preventing autoimmunity by distinguishing self from non-self RNA 8 . Similarly, the APOBEC family, which includes APOBEC1, APOBEC2, APOBEC3, and APOBEC4, serves diverse roles in lipid metabolism, viral defence, and genomic stability 9 . Specifically, APOBEC1 facilitates lipid transport by editing apolipoprotein B mRNA, a vital component of lipid metabolism mechanisms 10 . Moreover, members of the APOBEC3 subfamily, initially recognized for their role in combating retroviruses and mobile genetic elements by deaminating cytidines in single-stranded DNA, have recently been implicated in RNA editing. Environmental stimuli such as interferon activation, hypoxia, and cellular density changes can trigger APOBEC3A-induced C > U RNA editing in monocytes and other immune cells. These activities underscore their critical function in the innate immune defence, though they also pose risks for unintended mutations that might promote genomic instability and cancer 11 – 13 . Given the critical importance of RNA editing events, their detection using state-of-the-art technologies, such as next-generation sequencing, has become crucial. Fortunately, both A > I and C > U editing can be detected through sequencing of complementary DNA (cDNA) without the need for additional sample processing, yielding read-outs manifested as A > G and C > T mutations in the cDNA. Currently, highly sophisticated variant calling methods for genomic DNA have been well-established, with GATK and its 'best practices' from the Broad Institute being among the most prominent 14 , 15 . In contrast, the development of variant calling and analytical methods for RNA is still progressing, with no universally accepted standard yet in place. Several key challenges complicate the analysis of RNA editing events using next-generation sequencing. These include distinguishing genuine RNA edits from sequencing artefacts and DNA polymorphisms, circumventing errors introduced by special sequences such as repeats and homopolymer runs, and reliably comparing RNA editing profiles across different physiological conditions or disease states 11 , 16 , 17 . To tackle these challenges, several methodologies such as RVboost 18 , SNPiR 17 , VaDiR 19 , rMATS-DVR 20 , and JACUSA 21 , 22 have been developed. Each employs distinct analytical strategies to enhance the accuracy and reliability of R NA v ariants (RVs) or RNA editing sites detection, however, while each of these tools offers solutions to some of these challenges, a method that completely overcome all obstacles is still currently lacking. This paper elucidates the development and validation of Ca librated D ifferential R NA E diting S canner (CADRES), an analytical pipeline designed to detect highly reliable RNA editing events resulting from RNA base deamination. Compared to previous methods, CADRES not only filters out RVs that could potentially arise from Si ngle N ucleotide V ariant (SNV) or DNA mutations through joint DNA/RNA variant calling, but it also quantifies the level of RNA editing. This allows for the identification of D ifferential V ariants on R NA (DVRs) between multiple conditions. By demonstrating its capability to transform our understanding of RNA editing, our platform offers deeper insights into the complex roles of ADAR and APOBEC deaminases. Equipping researchers with these advanced tools, CADRES decodes the intricate relationship between RNA editing and disease, paving the way for novel diagnostic and therapeutic innovations. 2. Results 2.1. Implementation of the CADRES pipeline The CADRES pipeline is meticulously engineered to identify DVRs with exceptional precision. For a variant to be classified as a DVR, it must meet two stringent criteria. First, the variant must arise from RNA editing, rather than being a transcriptional artifact overlying a DNA mutation. Second, the variant must exhibit statistically significant differences in editing depth when comparing two distinct biological conditions, thereby pointing to genuine RNA editing sites that are crucial for studying the effects of biological variations on RNA editing (Fig. 1 ). This precise detection of DVRs by CADRES occurs in two primary phases, detailed in Fig. 2 : the R NA- D NA D ifference (RDD) phase and the R NA- R NA D ifference (RRD) phase. In the RDD phase, the pipeline systematically compares genomic DNA (gDNA) sequences obtained from W hole G enome S equencing (WGS) or W hole E xome S equencing (WES) against complementary DNA (cDNA) sequences from RNA sequencing (RNA-seq). This comparison is critical to filter out SNVs that could otherwise confound the identification of true RNA editing sites. In the RRD phase, the pipeline identifies DVRs by assessing statistical differences in variant depth across RNA-seq datasets from the two experimental conditions, each comprising replicates. Sites unaffected by experimental conditions are systematically excluded. The operation of the CADRES pipeline involves several meticulously orchestrated steps. Initial read mapping and alignment file preparation are conducted using Picard tools to ensure optimal data quality and alignment integrity 23 . The subsequent stage, termed 'boost recalibration,' involves joint DNA-RNA mutation calling using GATK4 MuTect2 24,25 . This process is vital for generating a library of de novo RNA editing sites under stringent selection criteria. This library, augmented by curated RNA editing sites from databases such as REDIportal 26 , serves as a 'known site' reference for the base quality score recalibration (BQSR) of RNA-seq data. During BQSR, these de novo RNA editing sites receive preferential treatment, minimizing the downgrading of quality scores due to sequencing artifacts. A comprehensive mutation calling is then re-performed on the recalibrated data, applying rigorous filters to eliminate potential artifacts and isolate bona fide RNA editing sites. The final analytical step utilizes the G eneralized L inear M ixed M odel (GLMM) in the rMATS statistical framework to sample the depth of reference and alternative alleles 24 , 25 . Only those RNA editing sites that demonstrate significant alterations in editing levels, as defined by predetermined statistical thresholds, are classified as DVRs. 2.2. In silico benchmark of detection methods for DVR callers To assess the efficacy of the CADRES pipeline and compare it with established methodologies capable of detecting DVRs, we conducted in silico benchmarking using simulated WGS and RNA-seq data, the detail of which is shown in Table 1 . We introduced two types of spike variants into the simulation: 50,000 SNVs present in both WGS and RNA-seq datasets, and 50,000 RVs exclusive to RNA-seq data. The frequencies of these RVs were adjusted across two conditions to ensure that 6,000 met the criteria for DVR designation by rMATS GLMM and JACUSA analysis, defining successful detection of DVRs as true positives by the benchmarked methods. Table 1 Description of in silico data WGS RNA-seq Library type 2×150 nt 2×100 nt, strand-specific Coverage/ read count 33× 16 million reads number of replicates 2 3 ~ 5 To simulate real-world scenarios more accurately, we also introduced spike DVR false positives by assigning approximately 2,000 SNVs with RNA variant frequency differences that met the DVR criteria in RNA-RNA-only detection methods. This setup established theoretical precision and accuracy ceilings of 0.75 and 0.98 for RNA-RNA-only callers, respectively, and 0.12 and 0.56 for RNA-DNA-only callers. A summary of the spiked variants are provided in Table 2 . Table 2 Detailed statistics of spiked sites use for benchmarking in silico data WGS RNA-seq Number spiked Mean AF† Number spiked Mean AF† ΔAF‡ SNV 48,000 0.05 ~ 0.5 48,000 0.05 ~ 0.5 ~ 0 False positive DVR 2,000 0.05 ~ 0.5 2,000 0.1 ~ 0.3 0.05 ~ 0.1 RV 0 0 ~ 44,000 0.1 ~ 0.3 ~ 0 DVR 0 0 ~ 6000 0.1 ~ 0.3 0.05 ~ 0.1 †: AF: A lternative allele F requency ‡: ΔAF = AF treatment - AF control The benchmarking extended to evaluating the impact of the 'boost recalibration' procedure implemented within the CADRES pipeline, compared to its omission. Concurrently, the same simulated datasets were analysed using JACUSA2, employing both RRD and RDD modes. To maximize the potential of DVR detection, a combined analysis of JACUSA2 was also conducted, where only variants identified by both RRD and RDD methods were retained, aligning with our stringent criteria for defining a DVR. Furthermore, the performances of VaDiR and rMATS-DVR methodologies were also evaluated. A comparative analysis of these methods is detailed in Table 3 , providing a comprehensive overview of their efficacy in identifying DVRs. Table 3 Comparison of methodologies evaluated in the benchmarking study BQSR Compares RRD Compares RDD Supports comparison between conditions Number of replicates supported Mutation calling engine Statistical analysis on variant depth Additional mapping artefact filters CADRES √ √ √ √ ≥ 2 GATK Mutect2 rMATS GLMM √ CADRES w/o boost recalibration √ √ √ √ ≥ 2 GATK Mutect2 rMATS GLMM √ JACUSA2 joint RDD-RRD √ √ √ ≥ 1 JACUSA Dirichlet-multinomial distribution √ JACUSA2 RRD mode √ √ ≥ 1 JACUSA Dirichlet-multinomial distribution √ rMATS-DVR √ √ √ ≥ 2 GATK UnifiedGenotyper rMATS GLMM JACUSA2 RDD mode √ √ ≥ 1 JACUSA Dirichlet-multinomial distribution √ VaDiR √ √ 1 GATK UnifiedGenotyper, GATK Mutect2 - √ Our analysis focused on how the number of replicates influenced the sensitivity and specificity of these methods, with each simulation being repeated three times. As depicted in Fig. 3 , while all methods demonstrated comparable true positive rates in detecting DVRs, distinct differences in precision and accuracy emerged among the different types of DVR callers. Methods that considered both RDD and RRD, such as CADRES and JACUSA2 joint RDD-RRD, achieved precision and accuracy scores that surpassed the limits set for RRD-only methods, indicating fewer spiked false positives. Notably, CADRES outperformed the JACUSA2 joint RDD-RRD mode in terms of precision by 10–15%, albeit at a slight cost of 3–5% in recall effectiveness. The boost recalibration procedure proved beneficial, enhancing both true positive rates (TPR) and precision, highlighting its importance in the CADRES pipeline. The inclusion of replicates generally improved benchmark scores, though the marginal benefits diminished when the number of replicates exceeded four. These findings substantiate that CADRES, particularly when applied with the boost recalibration technique and an optimal number of replicates, serves as an effective DVR detection tool, offering superior precision in identifying meaningful RNA editing events. 2.3. Real-World Evaluation of DVR Detection with Inducible A3B Cell Models To robustly assess the effectiveness of the CADRES pipeline, we employed a real-world biological model using lentiviral inducible cells engineered to express the DNA deaminase A3B (recombinant A3B-GFP) upon doxycycline induction. This setup enables direct comparison of A3B-induced samples with their non-induced counterparts, facilitating the detection of A3B-mediated C > U RNA editing. A3B was selected as a benchmark due to its well-documented role as an RNA editing enzyme and its distinct RNA editing motif (5’-UUCM, where M = C or A) 27 , 28 , different from its DNA motif (5’-TCW) 29 , 30 , which aids in validating RNA editing identifications. For this study, we utilized two distinct datasets: one from T-47D inducible cells, previously reported by our group 28 , and a newly generated dataset from SK-OV-3 cells. T-47D cells exhibit moderate endogenous A3B expression 31 ( Figure S1 ), providing a suitable background for detecting A3B-mediated DVRs, enhanced by the presence of A3B interactors. In contrast, SK-OV-3 cells, which show negligible endogenous A3B expression ( Figure S1 ), serve as a cleaner model to confirm DVRs specific to ectopically-expressed A3B. The concentration of doxycycline was meticulously controlled to induce A3B-GFP expression at levels 15–20 times higher than endogenous A3B in T-47D cells, while in SK-OV-3 cells, ectopic A3B-GFP levels were comparable to the endogenous A3B in T-47D cells. WGS was performed at over 30× coverage, and RNA-seq was conducted with four replicates for both induced and non-induced conditions. We evaluated the performance of CADRES alongside other methods including JACUSA2 joint RDD-RRD, rMATS-DVR, and VaDiR, across these datasets. As illustrated in Table 4 , all methods detected a higher number of A > G(I) and C > U variants compared to other variant types, suggesting successful identification of RNA editing events in T-47D cells. To confirm the biological relevance of identified DVRs, we analysed the minimum folding energy (MFE) of the sequences flanking DVRs using the RNAFold algorithm 32 . The MFE profiles of DVRs identified by all methodologies closely resembled those of established RNA editing sites, rather than SNVs or A3B-mediated DNA editing sites identified in a previous study 20 , 33 (GSE193225) (Fig. 4 a). This similarity indicates that the DVRs are indeed localized within RNA transcripts and are structured in a manner conducive to RNA folding. Further analysis of the nucleotide sequences flanking these editing sites showed that CADRES and JACUSA2 joint RDD-RRD methods performed best in enriching the ‘5’-UUCM’ motif for C > U DVRs (Fig. 4 b, S2), known to be preferentially targeted by A3B 27 . This motif enrichment was notably weaker in results from JACUSA2 RDD or RRD methods ( Figure S2 ), potentially due to the inclusion of a few non-A3B-mediated DVRs or SNVs. A noticeable overlap was observed between DVRs identified by rMATS-DVR and SNVs present in T-47D cells (Table 4 ). This overlap is expected as rMATS-DVR does not incorporate a mechanism to filter out SNVs. Given that rMATS-DVR and CADRES utilise similar analytical methodologies, we explored whether the inclusion of SNV contaminants might have influenced the outcomes of the rMATS GLMM model. As demonstrated in Figure S3 , the inclusion of SNVs in the rMATS-DVR analysis resulted in a significant proportion (33%) of C > U DVRs showing a decrease in alternative allele frequency following A3B-GFP induction, a pattern not observed in CADRES-derived DVRs. Typically, A3B induction is associated with an increase in alternative allele frequency at C > U DVRs, suggesting that the presence of SNVs may disrupt this expected pattern. Table 4 Comparison of number of DVRs identified by various methodologies in T-47D cells induced with A3B-GFP. CADRES JACUSA2 RDD JACUSA2 RRD JACUSA2 joint RDD-RRD VaDiR rMATs-DVR C > U DVRs 816 3399 1422 1048 1114 1540 A > G(I) DVRs 3764 82481 17753 14920 7485 26255 Other types of DVRs 29 11367 1791 396 8876 3230 C > U DVRs overlapping SNVs 0 2 13 0 0 878 A > G(I) DVRs overlapping SNVs 0 3 31 0 2 3150 Other types of DVRs 0 3 3 0 0 1099 In a further test using the SK-OV-3 cell model, which expresses ectopic A3B at a lower but biological relevant level, CADRES demonstrated a slightly reduced sensitivity in DVRS compared to JACUSA2 joint RDD-RRD (Table 5 , Fig. 5 a ) but achieved better enrichment for the specific editing motifs of A > G(I) and C > U. This finding underscores that while CADRES sacrifices some sensitivity to enhance specificity, it remains effective at detecting DVRs from biologically relevant alterations (Fig. 5 b). Collectively, these results affirm that the CADRES pipeline, when applied in conjunction with appropriate biological models and rigorous analytical techniques, provides highly specific and reliable detection of A3B-mediated DVRs across different experimental setups. Table 5 Comparison of number of DVRs identified by various methodologies in SK-OV-3 cells induced with A3B-GFP. CADRES JACUSA2 RDD JACUSA2 RRD JACUSA2 joint RDD-RRD C > U DVRs 442 2539 758 603 A > G(I) DVRs 165 63607 11293 9386 Other types of DVRs 3 29375 2179 1199 2.4. Validation of DVRs To further validate the DVRs identified by CADRES, we investigated the A3B binding at regions flanking these DVRs. Utilizing previously conducted e nhanced C ross- l inking and I mmuno p recipitation sequencing (eCLIP-seq) experiments that mapped A3B binding sites in RNA 28 (curated in GSE193225), we analysed the enrichment of A3B binding signals around the DVRs detected by CADRES, JACUSA2 RDD-RRD, rMATS-DVR and VaDiR methods. Consistent with expectations for a method demonstrating high accuracy, we observed significant enrichment of A3B binding signals at C > U DVRs derived from CADRES (Fig. 6 a). This binding pattern, particularly evident immediately upstream to the C > U DVRs (Fig. 6 b), confirms the presence of the A3B complex, which interferes with reverse transcription during library preparation. This validates that A3B specifically binds to RNA sites undergoing C > U editing, substantiating the successful identification of A3B-mediated DVRs. On the other hand, for JACUSA2-derived and rMATS-DVRs, enrichment of eCLIP-seq signal are only evident for high score DVRs, suggesting proportion of DVRs identified were not adjacent to A3B binding site (Fig. 6 a). Quantitative analysis revealed that CADRES-derived DVRs showed a higher level of A3B binding signal enrichment compared to those identified by JACUSA2-RDD-RRD, with CADRES showing approximately ~ 15–30% greater enrichment. To explore how this difference could arise we compared DVRs identified by CADRES and JACUSA2 joint RDD-RRD methods. As is shown in Fig. 7 , less than half of the DVRs from these methods intersect with each other, indicating significant different groups of DVRs being identified. Sequence motif analysis revealed that CADRES-derived DVRs display the distinct 5’-UUCM motif for A3B editing, while the JACUSA2-RDD-RRD-specific DVRs had a relatively less enrichment for A3B editing. This suggests that CADRES not only identifies A3B-derived C > U DVRs but does so with higher precision and specificity. 3. Discussion The CADRES pipeline developed in this study introduces a methodological enhancement for the detection of RNA editing events, particularly for DVRs. By seamlessly integrating sophisticated DNA/RNA variant calling with comprehensive statistical analyses, this tool refines the accuracy and specificity of RNA editing detection. The effective method for detection of DVRs not only provides a useful tool for probing the biological functions of RNA editing enzymes, but also facilitates detailed studies on how RNA editing is influenced by various experimental conditions, diseases, and therapeutic interventions. This capability allows researchers to dissect the complex mechanisms by which RNA editing impacts gene expression regulation, offering insights into disease progression and response to treatment. A pivotal attribute of the CADRES pipeline is its integration of both DNA-RNA and RNA-RNA comparative analyses. This dual comparison strategy effectively mitigates confounding factors from SNVs, which can introduce apparent RNA sequence variations during transcription. Additionally, the utilisation of the rMATS GLMM statistical model enables precise detection of differential RNA editing events 24 , 25 . Although the rigorous criteria employed by CADRES may result in a modest reduction in sensitivity compared to other methodologies, this is counterbalanced by a significant increase in precision. This high level of precision is critical in contexts where exact identification of DVRs is paramount, such as in functional studies of RNA editing enzymes or in clinical applications where precise biomarker identification can guide therapeutic decisions. In comparison to widely-used tools like JACUSA2, CADRES offers an enhanced capability for situations where precision is prioritized over sensitivity, providing researchers and clinicians with a robust tool for the accurate dissection of RNA editing dynamics. An essential enhancement in the CADRES pipeline is the tailored application of base BQSR for sequencing reads 25 . It is widely acknowledged among genomic researchers that BQSR is critical for all read alignments before mutation calling to address the common issue of pervasive sequence artifacts 34 , 35 . Typically, mismatches are considered artifacts by default, leading to their downgraded quality scores during recalibration, unless they coincide with pre-identified mutations which are less affected by the recalibration. While this strategy effectively isolates high-confidence SNVs, it can inadvertently diminish the detection sensitivity for RNA variants due to the incomplete cataloguing of known RNA editing sites. To mitigate this, the CADRES pipeline includes a 'boost recalibration' stage, involving a preliminary round of joint DNA-RNA mutation calling via GATK4 Mutect2. This process generates a list of tentative RVs that informs the subsequent BQSR of RNA-seq data. Our evaluations using both simulated and real datasets indicate that this modification slightly reduces the sensitivity loss typically associated with BQSR—less than 8% compared to methods that omit BQSR. This approach not only retains the method's precision by minimizing sequence artifacts but also ensures a balanced detection capability, crucial for accurate RNA editing analysis. The development of high-throughput sequencing technologies has substantially enhanced our capacity to study RNA editing. While the detection of A > I editing is relatively well-established due to its prevalence and biochemical detectability, the identification of C > U editing events remains challenging 36 . This is evidenced by the comparatively small repository of validated C > U editing sites, highlighting the inherent difficulties associated with accurately calling these modifications. In this study, we utilised an inducible APOBEC3B deaminase model to benchmark the performance of the CADRES pipeline in identifying C > U edits with high precision. Our findings suggest that CADRES offers an improvement in the detection of C > U editing events over existing methodologies. This capability underscores the importance of revisiting existing RNA-seq datasets, particularly those paired with whole-genome or whole-exome sequencing, to re-analyse them under more stringent criteria provided by CADRES. Looking forward, it would also be prudent to employ CADRES for the exploration of other less common types of RNA editing, such as U > C 37 , where traditional detection methods often fall short due to confounding factors like sequencing errors and SNVs. The CADRES methodology, while robust, presents avenues for further refinement to enhance its precision and adaptability, especially in complex biological samples. Our analysis indicates that despite employing WGS to filter out SNVs, residual shallow mutations continue to influence the accuracy of downstream analyses. This issue is particularly pronounced in samples with genomic heterogeneity, such as those derived from tumour tissues. An alternative approach could involve the use of WES, which typically offers higher sequencing depth and could more effectively mitigate the impact of these mutations. However, this substitution requires careful consideration of the genomic regions not covered by WES probes, which may also hold biological relevance. Furthermore, the precision of the CADRES pipeline could be substantially improved by expanding the database of known RNA editing sites. The accuracy of the BQSR process is directly proportional to the comprehensiveness of the RNA editing database utilized. As the scientific community continues to discover and catalogue more RNA editing events 26 , and as collaboration between researchers and database curators strengthens, we anticipate significant advancements in the fidelity of RNA editing detection. This evolution will enhance our ability to accurately profile RNA editing landscapes, paving the way for more nuanced biological insights and therapeutic strategies. In the development of the CADRES methodology, a deliberate trade-off was made to prioritize high predictive precision at the cost of a slight reduction in detection sensitivity. This approach utilizes stringent search criteria to ensure the accuracy of identified RNA editing sites. The question of whether this trade-off is justified merits consideration. We posit that the answer is affirmative, particularly in contexts such as RNA-editing enzyme research and the development of biomarkers, where the utility of a smaller number of highly precise RNA editing sites is evident. Furthermore, the burgeoning fields of deep learning and artificial intelligence, which rely heavily on the quality of input data for model training, stand to benefit significantly from the high precision of data provided by CADRES. While the sensitivity limitations of CADRES might be viewed as a drawback, these can potentially be mitigated by the inherent capacity of these computational methods to process large datasets efficiently. Looking ahead, we are optimistic about the integration of CADRES with advanced computational strategies, enhancing its utility and application in the rapidly evolving landscape of genomic research. 4. Materials and Methods 4.1 Cell culturing The T-47D human breast cancer cell line and SK-OV-3 human ovarian cancer cell line were obtained from the American Type Culture Collection (ATCC). Lenti-X 293T cells, used for lentivirus production, were obtained from Clontech. T-47D cells were cultured in RPMI-1640 medium (Gibco) supplemented with 10 µg/mL bovine insulin (Sigma), while SK-OV-3 cells were maintained in McCoy’s 5A medium (Gibco). Both media were supplemented with 10% v/v fetal bovine serum (FBS, PAA) and 0.5% v/v penicillin/streptomycin (pen/strep, Gibco). Lenti-X 293T cells were cultured in Dulbecco's Modified Eagle Medium (DMEM, Gibco) supplemented with 10% v/v FBS and 0.5% v/v pen/strep. All cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO2. Cell line authenticity was verified using short tandem repeat (STR) profiling, and all cell lines were regularly tested for mycoplasma contamination using PCR. 4.2 Generation of Stable Inducible Cell Lines Exponentially growing T-47D or SK-OV-3 cells were transfected with lentivirus in the presence of 10 µg/mL polybrene (Sigma) to facilitate viral entry. After a 48-hour transduction period, cells were selected with RPMI-1640 medium supplemented with 4 µg/mL puromycin (Gibco) to enrich for successfully transduced cells. Transduction efficiency was assessed by monitoring green fluorescent protein (GFP) expression and was further confirmed by quantitative reverse transcription PCR (qRT-PCR) after induction with doxycycline. To maintain the lentiviral cassettes in the T-47D and SK-OV-3 cell lines, cells were cultured continuously in medium containing 4 µg/mL puromycin. 4.3 Whole-genome and RNA sequencing Cells were harvested using trypsin digestion. Genomic DNA was extracted using a Qiagen Genomic-tips 500/G kit. Sequencing libraries for WGS with DNA nanoball (DNB) technology were constructed (Beijing Genomics Institute (BGI) Inc., Hong Kong). Sequencing was carried out using BGISEQ-500 sequencer with a mean sequencing coverage of greater than 30× for each of the samples using 2×150 bp configuration. Sequence reads from WGS was aligned to GRCh38 genome assembly using Burrow-Wheelers aligner 38 . Total RNA were extracted from cells using the MagNA pure 96 platform (Roche), and RNA integrity number (RIN) was determined by BioAnalyzer 2100 (Agilent). RNA libraries were constructed and sequenced using poly-dT enrichment or ribosomal RNA depletion sample preparation methods in conjunction with DNB sequencing technology on a BGISEQ-500 instrument (BGI). Raw reads were aligned to GRCh38 genome assembly and GENCODE GRCh38.p13 annotation using STAR 39 , 40 , and then processed with SAMtools 41 . For RNA-seq in this study, four biological replicates were conducted. 4.4. Detection of DVRs using CADRES pipeline Data preprocessing: Preprocessing of sequencing alignments were carried out using Picard tools 19 , 20 . The sequence alignment for WGS and RNA-seq, both in .bam format, were assigned with appropriate read groups. For samples sequenced with BGISEQ-500 platform, ‘COMPLETE’ (Complete Genomics) was chosen as sequence platform (‘RGPL = COMPLETE’). Duplicate sequence reads were then marked. For RNA-seq alignments, reads containing ‘N’ CIGAR strings were split, while for WGS alignments, base quality score was recalibrated using recommended procedures provided by ‘GATK best practice workflow’ introduced by the Broad Institute, where dbSNP150 database 42 , 43 resource were used as ‘known’ sites. Boost recalibration for RNA variants: This process involves joint variant calling and processing using the Mutect2 programme from GATK4 package 24 , 25 . For each of the conditions, the RNA-seq data was defined as ‘Tumour’ samples and WGS data was defines as is corresponding ‘Normal’ control. The resultant variant calls were then selected using default Mutect2 filters, and the ‘PASS’ variants were retained. The ‘PASS’ variants for both conditions were then collated as putative RNA variants, and then combined with known RNA editing sites from REDIportal 26 . The resulting list of variants were the used as ‘known’ sites for the base recalibration procedure of each of the RNA-seq alignment file. Joint DNA/RNA mutation calling for RNA variants: A second round of joint mutation calling were performed using Mutect2, with RNA-seq alignments as ‘Tumour’ samples and WGS alignments as ‘Normal’ samples, yielding a list of variants that undergoes filtering using standard Mutect2 filters. During this filtering step, maximum allowed mutations within one sequencing read was set to 4 instead of 2 to accommodate for the possible clustered nature for certain RNA editing enzymes, such as APOBEC3B which contains two deaminase domains form a homodimer. In addition to the default Mutect2 filters, two filters from the previously developed SNPiR pipeline were adopted to remove variants at the end of homopolymer runs and variants within multiple mapped sequences 17 . This was achieved by scripts previously by Piskol et al., in conjunction with the pBLAT programme 44 . After these filtering processes, the resulting variants are defined as RNA variants. Sampling of allele depth for RNA variants: The aligned sequence reads for each sample in both conditions were separated by reference genomic strand using SAM flag string. For each of the RNA variants, the depth for reference and alternative allele was counted in a strand-specific manner using SAMtools ‘mpileup’ programme, which generates an allele depth table of RVs for each of the samples. rMATS statistical analysis: rMATS statistical analysis were performed using the previously developed rMATS-DVR programme 20 . Briefly, the allele depth table were subjected to rMATS (replicate Multivariate Analysis of Transcript Splicing) statistical analysis, which uses a generalized linear mixed model (GLMM) to simultaneously account for the RNA-seq estimation uncertainty in the alternative allele frequency as influenced by sequencing coverage in individual samples, and the variability in isoform ratios among replicates 45 . This process yields a probability and false discovery rate (FDR) for a RV being a DVR. In this study, to be identified as a DVR the FDR requires less than 0.05, with minimal sequencing depth needs to be at least 10. The resulting DVR were then annotated with RefSeq. 4.5 JACUSA2 joint RDD and RRD calling Aligned files were processed using MarkDuplicates programme in Picard tools. For the processing of data, developer’s instructions were followed so that experimental design ( i.e. replications within condition groups) was recognised by JACUSA2 programme. Both RDD and RRD analyses were conducted independently. The outputs from these analyses were further processed using the JACUSA2helper package in R. Variant filtering was performed using the default settings of the programme. For the joint RDD-RRD analysis, only variants that met the criteria for both RDD and RRD were included in the subsequent analysis. 4.6 Benchmark of DVR detection methods using simulated data Simulated sequencing data were generated using the same method as reported by Piechotta et al. 22 with the following exceptions. First, mutation-free WGS data was generated using ReSeq using developers’ recommended setting, and RNA-seq data was generated using RESM with T-47D data as template for expression. Second, variant frequencies for SNVs were sampled from a previous study using publicly deposited data, which was applied to both WGS and RNA-seq samples. Third, two types of variants were spiked into mutation-free simulated data, i.e. 50,000 SNVs and 50,000 RVs, as is summarised in Table 1 . Prior to benchmark analysis on simulated data, statistical analysis on spiked variants using rMATS GLMM and/or JACUSA methods were conducted to distinguish DVRs from other RVs, as well as calculating the theoretical best for variant calling methods. The identification of DVRs were defined as true positive in the following analysis. 4.7 Benchmark of DVR detection using inducible A3B deaminase models For inducible expression of the lentiviral-encoded proteins, T-47D and SK-OV-3 cells cells were exposed to culture media supplemented with 100 ng/mL doxycycline (Sigma) to trigger protein expression. After 72 hours of induction, the expression of ectopic A3B were monitored by RT-qPCR, ensuring that the ectopic A3B is 10–20 fold expression to its endogenous counterpart. The cell samples were subjected to WGS and RNA-seq. 4.8 DVR motif analysis Sequences flanking the identified DVRs were extracted using the GRChg38 reference sequence. The nucleotide composition around these sites was visualized using Weblogo3 to analyse sequence conservation and motif prevalence 46 . 4.9 RNA folding analysis RNA secondary structure predictions were performed using the RNAFold programme 32 , supported by the LncFinder package from R 47 , which aids in identifying potential long non-coding RNA structures. For this analysis, a window of ± 100 base pairs around each identified SNV or RNA editing site was considered. To ensure a robust comparison, SNVs were sampled from the dbSNP database 42 , 43 with a random selection of 100,000 variants. Control sequences were selected based on RNA-seq data with a read depth greater than 10, using BEDtools to isolate relevant gene segments. These segments were then randomly sampled (n = 1,000) to match the sequence motifs of interest 48 . The folding conditions for the RNAFold analysis were set to physiological conditions (37°C, with a salt concentration of 1.0 M and no lonely pairs, -T 37 -salt 1.0 -d 2 --noLP), which are crucial for accurately modelling RNA folding dynamics in a cellular environment. 4.10 Visualisation of eCLIP-seq data eCLIP-seq data were converted to bigwig file format, which denotes fold change of immunoprecipitated signal over IgG size-matched control. The files were then visualised using Seqplots ( https://przemol.github.io/seqplots/ ) programme. Declarations Author Contributions: Conceptualisation, C.Z. and X.L.; methodology, C.Z.; software, J.S.; investigation, J.S. and C.Z.; writing, C.Z. and J.S.; visualisation, C.Z. and J.S.; supervision, X.L.; project administration, X.L.; funding acquisition, C.Z. and X.L. All authors have read and agreed to the published version of the manuscript. Data Availability Statement: CADRES is coded by a combination of Python and Shell script languages, which is available at GitHub (https://github.com/junsun-hash/CADRES). Sequencing data related to this study is available at GEO with accession numbers (GSE193225, GSExxxxx). Acknowledgments: The authors thank the funding support from Shanghai Institute of Biological Products. C.Z. was sponsored by Shanghai Pujiang Programme (22PJD104), and Science and Technology Commission of Shanghai (23S11901100). Conflicts of Interest: J.S., C.Z., and X.L. are employees of Shanghai Institute of Biological Products, an entity presently engaged in the commercial development of therapeutic biologics. References Qiu, L., Jing, Q., Li, Y. & Han, J. RNA modification: mechanisms and therapeutic targets. Mol Biomed 4, 25 (2023). Eisenberg, E. & Levanon, E. Y. A-to-I RNA editing - immune protector and transcriptome diversifier. Nat Rev Genet 19, 473–490 (2018). Slotkin, W. & Nishikura, K. Adenosine-to-inosine RNA editing and human disease. Genome Med 5, 105 (2013). Pecori, R., Di Giorgio, S., Paulo Lorenzo, J. & Nina Papavasiliou, F. Functions and consequences of AID/APOBEC-mediated DNA and RNA deamination. Nat Rev Genet 23, 505–518 (2022). Shi, H., Chai, P., Jia, R. & Fan, X. Novel insight into the regulatory roles of diverse RNA modifications: Re-defining the bridge between transcription and translation. Mol Cancer 19, 78 (2020). Satam, H. et al. Next-Generation Sequencing Technology: Current Trends and Advancements. Biology (Basel) 12 (2023). Walkley, C. R. & Li, J. B. Rewriting the transcriptome: adenosine-to-inosine RNA editing by ADARs. Genome Biol 18, 205 (2017). Hu, S. B. & Li, J. B. RNA editing and immune control: from mechanism to therapy. Curr Opin Genet Dev 86, 102195 (2024). Salter, J. D., Bennett, R. P. & Smith, H. C. The APOBEC Protein Family: United by Structure, Divergent in Function. Trends Biochem Sci 41, 578–594 (2016). Blanc, V. & Davidson, N. O. APOBEC-1-mediated RNA editing. Wiley Interdiscip Rev Syst Biol Med 2, 594–602 (2010). Sharma, S. et al. APOBEC3A cytidine deaminase induces RNA editing in monocytes and macrophages. Nat Commun 6, 6881 (2015). Baysal, B. E. et al. Hypoxia-inducible C-to-U coding RNA editing downregulates SDHB in monocytes. PeerJ 1, e152 (2013). Alqassim, E. Y. et al. RNA editing enzyme APOBEC3A promotes pro-inflammatory M1 macrophage polarization. Commun Biol 4, 102 (2021). Van der Auwera, G. A. et al. From FastQ data to high confidence variant calls: the Genome Analysis Toolkit best practices pipeline. Curr Protoc Bioinformatics 43, 11 10 11–11 10 33 (2013). Best Practices for Variant Calling with the GATK , ( Bahn, J. H. et al. Accurate identification of A-to-I RNA editing in human by transcriptome sequencing. Genome Res 22, 142–150 (2012). Piskol, R., Ramaswami, G. & Li, J. B. Reliable identification of genomic variants from RNA-seq data. Am J Hum Genet 93, 641–651 (2013). Wang, C. et al. RVboost: RNA-seq variants prioritization using a boosting method. Bioinformatics 30, 3414–3416 (2014). Neums, L. et al. VaDiR: an integrated approach to Variant Detection in RNA. Gigascience 7, 1–13 (2018). Wang, J., Pan, Y., Shen, S., Lin, L. & Xing, Y. rMATS-DVR: rMATS discovery of differential variants in RNA. Bioinformatics 33, 2216–2217 (2017). Piechotta, M., Naarmann-de Vries, I. S., Wang, Q., Altmuller, J. & Dieterich, C. RNA modification mapping with JACUSA2. Genome Biol 23, 115 (2022). Piechotta, M., Wyler, E., Ohler, U., Landthaler, M. & Dieterich, C. JACUSA: site-specific identification of RNA editing events from replicate sequencing data. BMC Bioinformatics 18, 7 (2017). Picard Tools. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 20, 1297–1303 (2010). Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat Biotechnol 31, 213–219 (2013). Mansi, L. et al. REDIportal: millions of novel A-to-I RNA editing events from thousands of RNAseq experiments. Nucleic Acids Res 49, D1012-D1019 (2021). Alonso de la Vega, A. et al. Acute expression of human APOBEC3B in mice results in RNA editing and lethality. Genome Biol 24, 267 (2023). Zhang, C. L., Y.; Wang, M.; Chen, B.; Xiong, F.; Mitsopoulos, C.; Rosanesse, O.; Clarke, P. Characterisation of APOBEC3B-Mediated RNA Editing in Breast Cancer Cells Reveals Regulatory Roles of NEAT1 and MALAT1 lncRNAs. Research Gate (2023). Burns, M. B. et al. APOBEC3B is an enzymatic source of mutation in breast cancer. Nature 494, 366–370 (2013). Burns, M. B., Temiz, N. A. & Harris, R. S. Evidence for APOBEC3B mutagenesis in multiple human cancers. Nat Genet 45, 977–983 (2013). Periyasamy, M. et al. APOBEC3B-Mediated Cytidine Deamination Is Required for Estrogen Receptor Action in Breast Cancer. Cell Rep 13, 108–121 (2015). Gruber, A. R., Lorenz, R., Bernhart, S. H., Neubock, R. & Hofacker, I. L. The Vienna RNA websuite. Nucleic Acids Res 36, W70-74 (2008). Zhang, C. et al. R-loop editing by DNA cytosine deaminase APOBEC3B determines the activity of estrogen receptor enhancers. bioRxiv (2022). Zverinova, S. & Guryev, V. Variant calling: Considerations, practices, and developments. Hum Mutat 43, 976–985 (2022). Koboldt, D. C. Best practices for variant calling in clinical sequencing. Genome Med 12, 91 (2020). Lerner, T. et al. C-to-U RNA Editing: From Computational Detection to Experimental Validation. Methods Mol Biol 2181, 51–67 (2021). Knoop, V. C-to-U and U-to-C: RNA editing in plant organelles and beyond. J Exp Bot 74, 2273–2294 (2023). Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009). Harrow, J. et al. GENCODE: the reference human genome annotation for The ENCODE Project. Genome Res 22, 1760–1774 (2012). Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013). Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10 (2021). Smigielski, E. M., Sirotkin, K., Ward, M. & Sherry, S. T. dbSNP: a database of single nucleotide polymorphisms. Nucleic Acids Res 28, 352–355 (2000). Sherry, S. T., Ward, M. & Sirotkin, K. dbSNP-database for single nucleotide polymorphisms and other classes of minor genetic variation. Genome Res 9, 677–679 (1999). Wang, M. & Kong, L. pblat: a multithread blat algorithm speeding up aligning sequences to genomes. BMC Bioinformatics 20, 28 (2019). Shen, S. et al. rMATS: robust and flexible detection of differential alternative splicing from replicate RNA-Seq data. Proc Natl Acad Sci U S A 111, E5593-5601 (2014). Crooks, G. E., Hon, G., Chandonia, J. M. & Brenner, S. E. WebLogo: a sequence logo generator. Genome Res 14, 1188–1190 (2004). Han, S. et al. LncFinder: an integrated platform for long non-coding RNA identification utilizing sequence intrinsic composition, structural information and physicochemical property. Brief Bioinform 20, 2009–2027 (2019). Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010). Additional Declarations Competing interest reported. J.S., C.Z., and X.L. are employees of Shanghai Institute of Biological Products, an entity presently engaged in the commercial development of therapeutic biologics. Supplementary Files SupplementalList1.xlsx Supplemental List 1:DVRs identified by various DVR callers in A3B-inducible T-47D cells (related to Figure 4), SupplementalList2.xlsx Supplemental List 2: DVRs identified by various DVR callers in A3B-inducible SK-OV-3 cells (related to Figure 5). SupplementalFiguresS1S3.pptx Figure S1-S4 Cite Share Download PDF Status: Published Journal Publication published 04 Jun, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 10 Jan, 2025 Reviews received at journal 16 Dec, 2024 Reviewers agreed at journal 02 Dec, 2024 Reviewers agreed at journal 29 Oct, 2024 Reviews received at journal 27 Sep, 2024 Reviewers agreed at journal 18 Sep, 2024 Reviewers invited by journal 02 Sep, 2024 Editor assigned by journal 24 Aug, 2024 Editor invited by journal 20 Aug, 2024 Submission checks completed at journal 14 Aug, 2024 First submitted to journal 04 Aug, 2024 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-4855669","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":335688340,"identity":"0770e272-06f6-43d9-b30d-969103aa2ba3","order_by":0,"name":"Jun Sun","email":"","orcid":"","institution":"Shanghai Institute of Biological Products","correspondingAuthor":false,"prefix":"","firstName":"Jun","middleName":"","lastName":"Sun","suffix":""},{"id":335688341,"identity":"4e3df37e-126a-49ea-b52a-08ad7a03830d","order_by":1,"name":"Chi Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYDACCcYGBiBiYGBmPsDwwIAkLexsCQwJxGkBYrAWfh4DhgRidPDPbm5g/LnDLk/emefjh4SCw4nbGZgfPrqBz5I7BxsYJM8kFxse5t0skWBwOHFnA5uxcQ4eLQYSiQ0Mhm3MiRubeTeAtWw4wMMmTVBLYls9UAvP4x/EaznYdjhxPjMPG3G2SNxIbDjY2HY8cQMzm5lFgkG68YbDBPzCPyP94cOfbdWJ8/sPP77x4Y+17IbjzQ8f49MCAgfALgSTDM3AZEBAORzIN4CpOmLVj4JRMApGwQgCAL9AUVR/Hk2/AAAAAElFTkSuQmCC","orcid":"","institution":"Shanghai Institute of Biological Products","correspondingAuthor":true,"prefix":"","firstName":"Chi","middleName":"","lastName":"Zhang","suffix":""},{"id":335688342,"identity":"de9d1487-79db-4cc3-8efa-6b4591fac2e9","order_by":2,"name":"Xiuling Li","email":"","orcid":"","institution":"Shanghai Institute of Biological Products","correspondingAuthor":false,"prefix":"","firstName":"Xiuling","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2024-08-04 06:53:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4855669/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4855669/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-04957-7","type":"published","date":"2025-06-04T15:57:31+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":81506445,"identity":"32ab47fc-556b-4d43-b996-ea45b7938042","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":85468,"visible":true,"origin":"","legend":"\u003cp\u003eDefinition of \u003cu\u003eD\u003c/u\u003eifferential \u003cu\u003eV\u003c/u\u003eariant on \u003cu\u003eR\u003c/u\u003eNA (DVR) in this study and potential confounding factors affecting their detection.\u003c/p\u003e","description":"","filename":"Slide1.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/788a8fe170999c3fe2ef93a6.png"},{"id":81506446,"identity":"86ebec87-024e-4c16-8a6f-919d43ee7d29","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":52254,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic for the \u003cu\u003eCa\u003c/u\u003elibrated \u003cu\u003eD\u003c/u\u003eifferential \u003cu\u003eR\u003c/u\u003eNA \u003cu\u003eE\u003c/u\u003editing \u003cu\u003eS\u003c/u\u003ecanner (CADRES) analysis pipeline.\u003c/p\u003e","description":"","filename":"Slide2.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/f903518921b30f9ff1c92f6b.png"},{"id":81508309,"identity":"201f5eb0-86b3-45ae-a1ba-11f53949e51b","added_by":"auto","created_at":"2025-04-28 05:38:35","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":53200,"visible":true,"origin":"","legend":"\u003cp\u003eBenchmark results for \u003cem\u003ein silico\u003c/em\u003e DVR detection. Average values of n = 3 simulations are shown.\u003c/p\u003e","description":"","filename":"Slide3.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/c3ec45b6709a72521ee61351.png"},{"id":81506453,"identity":"dca7daa0-598a-473c-a332-6e888be083b7","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":87293,"visible":true,"origin":"","legend":"\u003cp\u003eCharacteristics of DVRs identified in A3B-inducible T-47D cells. \u003cstrong\u003e(a) \u003c/strong\u003eDensity plots depicting the MFE derived from RNAFold analyses for DVRs. \u003cstrong\u003e(b) \u003c/strong\u003eSequence logos representing the nucleotide frequency surrounding the editing sites. For (a) and (b), 10,000 randomly sampled A\u0026gt;I editing sites from REDIportal and data for A3B-mediated DNA editing sites curated in dataset GSE193225 were used.\u003c/p\u003e","description":"","filename":"Slide4.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/f7eeb677f1b6b8b3152874af.png"},{"id":81506452,"identity":"a7f98e4b-d35e-4a5d-91f5-531a986bcc06","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":63812,"visible":true,"origin":"","legend":"\u003cp\u003eCharacteristics of DVRs identified in A3B-inducible SK-OV-3 cells. \u003cstrong\u003e(a)\u003c/strong\u003e Density plots depicting the MFE derived from RNAFold analyses for DVRs. \u003cstrong\u003e(b) \u003c/strong\u003eSequence logos representing the nucleotide frequency surrounding the editing sites.\u003c/p\u003e","description":"","filename":"Slide5.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/7b0a9997ede871e8b27794d3.png"},{"id":81506449,"identity":"cdac01f4-b540-4293-be72-22dd24855670","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":356289,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of DVRs using enhanced \u003cu\u003eC\u003c/u\u003eross-\u003cu\u003el\u003c/u\u003einking and \u003cu\u003eI\u003c/u\u003emmuno\u003cu\u003ep\u003c/u\u003erecipitation sequencing (eCLIP-seq) data. (a) Heatmaps depicting the enrichment of eCLIP-seq signals, which indicate A3B binding at locations surrounding the identified DVRs across different DVR calling methods. Each row in the heatmaps represents a specific C\u0026gt;U DVR, with DVRs sorted according to the criteria specified in the heatmaps. (b) Profiles of eCLIP-seq signals for C\u0026gt;U DVRs identified by the CADRES and JACUSA2 methods. Lines represent the mean values, and shaded areas indicate 95% confidence intervals for these signals.\u003c/p\u003e","description":"","filename":"Slide6.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/d77f97ddf8047f462b1eccb7.png"},{"id":81508316,"identity":"0e9a0827-62ae-42b6-959e-4d783a5578e2","added_by":"auto","created_at":"2025-04-28 05:38:35","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":44002,"visible":true,"origin":"","legend":"\u003cp\u003eComparative Analysis of C\u0026gt;U DVRs Identified by CADRES and the joint RDD-RRD method by JACUSA2. Venn diagram showing the overlap of C\u0026gt;U DVRs identified using the CADRES pipeline and the joint RDD-RRD method implemented by JACUSA2. Additionally, sequence motif analyses are provided for the DVRs located in the specific sections indicated in the diagram.\u003c/p\u003e","description":"","filename":"Slide7.png","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/35f96bec24c9f5e1be2853ef.png"},{"id":84243134,"identity":"7e5b5ea6-f5aa-45c3-b376-54081dc6090b","added_by":"auto","created_at":"2025-06-09 16:12:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1631535,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/6f037cd6-6d6e-4d77-8b30-02b6dcc8e4d8.pdf"},{"id":81506456,"identity":"18f673d7-f225-4bbf-8aae-3813744942d2","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":27448294,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplemental List 1:\u003c/strong\u003eDVRs identified by various DVR callers in A3B-inducible T-47D cells (related to Figure 4),\u003c/p\u003e","description":"","filename":"SupplementalList1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/8e603f22deed739af7be0c93.xlsx"},{"id":81509713,"identity":"58fde9a1-d96e-47f5-9b5b-7078b96ff1e7","added_by":"auto","created_at":"2025-04-28 06:02:23","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":7589129,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplemental List 2: \u003c/strong\u003eDVRs identified by various DVR callers in A3B-inducible SK-OV-3 cells (related to Figure 5)\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"SupplementalList2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/cc0a4070de00d90347319577.xlsx"},{"id":81506459,"identity":"e69b7c38-2536-49a7-98bc-3c7c0ecdf139","added_by":"auto","created_at":"2025-04-28 05:30:35","extension":"pptx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":3251340,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure S1-S4\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"SupplementalFiguresS1S3.pptx","url":"https://assets-eu.researchsquare.com/files/rs-4855669/v1/d00f8b31f67b39a44fc08ea9.pptx"}],"financialInterests":"Competing interest reported. J.S., C.Z., and X.L. are employees of Shanghai Institute of Biological Products, an entity presently engaged in the commercial development of therapeutic biologics.","formattedTitle":"CADRES: An Analytical Pipeline for Precise Identification of Differential RNA Editing Events Across Varied Biological Conditions","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003e \u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eRNA editing is a crucial post-transcriptional modification that enables organisms to dynamically alter transcript sequences without modifying the underlying genomic DNA \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. This process is fundamental in enhancing proteomic diversity and exerting regulatory effects on gene expression across higher eukaryotes \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. At the core of this mechanism are enzymatic activities such as the conversion of adenine to inosine (A \u0026gt; I) and cytidine to uracil (C \u0026gt; U), which are recognized during translation as guanosine and uridine, respectively. These modifications can significantly alter the amino acid sequences encoded by the genomic DNA, potentially impacting protein function and cellular behaviour \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Beyond these well-known types, other RNA editing mechanisms, including those involving methylation and pseudouridylation, also play critical roles in cellular response to environmental changes and disease pathogenesis \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. The expanding understanding of RNA editing, propelled by advancements in sequencing technologies \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, underscores its importance in biological regulation and disease development.\u003c/p\u003e \u003cp\u003eAmong the most significant contributors to RNA editing are the ADAR (\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eA\u003c/span\u003edenosine \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eeaminases \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eA\u003c/span\u003ecting on \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA) and APOBEC (Apolipoprotein B mRNA Editing Enzyme, Catalytic Polypeptide-like) families of deaminases \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. ADAR enzymes, comprising ADAR1, ADAR2, and ADAR3, are essential for A \u0026gt; I editing and play a critical role in normal cellular function. They modulate the coding potential and expression of neurotransmitter receptors and ion channels, crucial for maintaining synaptic plasticity and proper neuronal signalling \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Beyond their role in the brain, ADAR enzymes are involved in the immune system, where they edit RNA transcripts encoding interferon-inducible proteins, thus preventing autoimmunity by distinguishing self from non-self RNA \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Similarly, the APOBEC family, which includes APOBEC1, APOBEC2, APOBEC3, and APOBEC4, serves diverse roles in lipid metabolism, viral defence, and genomic stability \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Specifically, APOBEC1 facilitates lipid transport by editing apolipoprotein B mRNA, a vital component of lipid metabolism mechanisms \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Moreover, members of the APOBEC3 subfamily, initially recognized for their role in combating retroviruses and mobile genetic elements by deaminating cytidines in single-stranded DNA, have recently been implicated in RNA editing. Environmental stimuli such as interferon activation, hypoxia, and cellular density changes can trigger APOBEC3A-induced C \u0026gt; U RNA editing in monocytes and other immune cells. These activities underscore their critical function in the innate immune defence, though they also pose risks for unintended mutations that might promote genomic instability and cancer \u003csup\u003e\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e–\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eGiven the critical importance of RNA editing events, their detection using state-of-the-art technologies, such as next-generation sequencing, has become crucial. Fortunately, both A \u0026gt; I and C \u0026gt; U editing can be detected through sequencing of complementary DNA (cDNA) without the need for additional sample processing, yielding read-outs manifested as A \u0026gt; G and C \u0026gt; T mutations in the cDNA. Currently, highly sophisticated variant calling methods for genomic DNA have been well-established, with GATK and its 'best practices' from the Broad Institute being among the most prominent \u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. In contrast, the development of variant calling and analytical methods for RNA is still progressing, with no universally accepted standard yet in place. Several key challenges complicate the analysis of RNA editing events using next-generation sequencing. These include distinguishing genuine RNA edits from sequencing artefacts and DNA polymorphisms, circumventing errors introduced by special sequences such as repeats and homopolymer runs, and reliably comparing RNA editing profiles across different physiological conditions or disease states \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. To tackle these challenges, several methodologies such as RVboost \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, SNPiR \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e, VaDiR \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, rMATS-DVR \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, and JACUSA \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e have been developed. Each employs distinct analytical strategies to enhance the accuracy and reliability of \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ev\u003c/span\u003eariants (RVs) or RNA editing sites detection, however, while each of these tools offers solutions to some of these challenges, a method that completely overcome all obstacles is still currently lacking.\u003c/p\u003e \u003cp\u003eThis paper elucidates the development and validation of \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eCa\u003c/span\u003elibrated \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifferential \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eE\u003c/span\u003editing \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eS\u003c/span\u003ecanner (CADRES), an analytical pipeline designed to detect highly reliable RNA editing events resulting from RNA base deamination. Compared to previous methods, CADRES not only filters out RVs that could potentially arise from \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eSi\u003c/span\u003engle \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eN\u003c/span\u003eucleotide \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eV\u003c/span\u003eariant (SNV) or DNA mutations through joint DNA/RNA variant calling, but it also quantifies the level of RNA editing. This allows for the identification of \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifferential \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eV\u003c/span\u003eariants on \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA (DVRs) between multiple conditions. By demonstrating its capability to transform our understanding of RNA editing, our platform offers deeper insights into the complex roles of ADAR and APOBEC deaminases. Equipping researchers with these advanced tools, CADRES decodes the intricate relationship between RNA editing and disease, paving the way for novel diagnostic and therapeutic innovations.\u003c/p\u003e \u003c/div\u003e "},{"header":"2. Results","content":"\u003ch2\u003e2.1. Implementation of the CADRES pipeline\u003c/h2\u003e\u003cp\u003eThe CADRES pipeline is meticulously engineered to identify DVRs with exceptional precision. For a variant to be classified as a DVR, it must meet two stringent criteria. First, the variant must arise from RNA editing, rather than being a transcriptional artifact overlying a DNA mutation. Second, the variant must exhibit statistically significant differences in editing depth when comparing two distinct biological conditions, thereby pointing to genuine RNA editing sites that are crucial for studying the effects of biological variations on RNA editing (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis precise detection of DVRs by CADRES occurs in two primary phases, detailed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e: the \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA-\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eNA \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifference (RDD) phase and the \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA-\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifference (RRD) phase. In the RDD phase, the pipeline systematically compares genomic DNA (gDNA) sequences obtained from \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eW\u003c/span\u003ehole \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eG\u003c/span\u003eenome \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eS\u003c/span\u003eequencing (WGS) or \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eW\u003c/span\u003ehole \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eE\u003c/span\u003exome \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eS\u003c/span\u003eequencing (WES) against complementary DNA (cDNA) sequences from RNA sequencing (RNA-seq). This comparison is critical to filter out SNVs that could otherwise confound the identification of true RNA editing sites. In the RRD phase, the pipeline identifies DVRs by assessing statistical differences in variant depth across RNA-seq datasets from the two experimental conditions, each comprising replicates. Sites unaffected by experimental conditions are systematically excluded.\u003c/p\u003e\u003cp\u003eThe operation of the CADRES pipeline involves several meticulously orchestrated steps. Initial read mapping and alignment file preparation are conducted using Picard tools to ensure optimal data quality and alignment integrity \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. The subsequent stage, termed 'boost recalibration,' involves joint DNA-RNA mutation calling using GATK4 MuTect2 \u003csup\u003e24,25\u003c/sup\u003e. This process is vital for generating a library of \u003cem\u003ede novo\u003c/em\u003e RNA editing sites under stringent selection criteria. This library, augmented by curated RNA editing sites from databases such as REDIportal \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, serves as a 'known site' reference for the base quality score recalibration (BQSR) of RNA-seq data. During BQSR, these de novo RNA editing sites receive preferential treatment, minimizing the downgrading of quality scores due to sequencing artifacts. A comprehensive mutation calling is then re-performed on the recalibrated data, applying rigorous filters to eliminate potential artifacts and isolate \u003cem\u003ebona fide\u003c/em\u003e RNA editing sites. The final analytical step utilizes the \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eG\u003c/span\u003eeneralized \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eL\u003c/span\u003einear \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eM\u003c/span\u003eixed \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eM\u003c/span\u003eodel (GLMM) in the rMATS statistical framework to sample the depth of reference and alternative alleles \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Only those RNA editing sites that demonstrate significant alterations in editing levels, as defined by predetermined statistical thresholds, are classified as DVRs.\u003c/p\u003e\u003ch2\u003e2.2. In silico benchmark of detection methods for DVR callers\u003c/h2\u003e\u003cp\u003eTo assess the efficacy of the CADRES pipeline and compare it with established methodologies capable of detecting DVRs, we conducted \u003cem\u003ein silico\u003c/em\u003e benchmarking using simulated WGS and RNA-seq data, the detail of which is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. We introduced two types of spike variants into the simulation: 50,000 SNVs present in both WGS and RNA-seq datasets, and 50,000 RVs exclusive to RNA-seq data. The frequencies of these RVs were adjusted across two conditions to ensure that 6,000 met the criteria for DVR designation by rMATS GLMM and JACUSA analysis, defining successful detection of DVRs as true positives by the benchmarked methods.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescription of \u003cem\u003ein silico\u003c/em\u003e data\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWGS\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRNA-seq\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLibrary type\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2×150 nt\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2×100 nt, strand-specific\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoverage/ read count\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33×\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u0026nbsp;million reads\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003enumber of replicates\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3 ~ 5\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eTo simulate real-world scenarios more accurately, we also introduced spike DVR false positives by assigning approximately 2,000 SNVs with RNA variant frequency differences that met the DVR criteria in RNA-RNA-only detection methods. This setup established theoretical precision and accuracy ceilings of 0.75 and 0.98 for RNA-RNA-only callers, respectively, and 0.12 and 0.56 for RNA-DNA-only callers. A summary of the spiked variants are provided in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDetailed statistics of spiked sites use for benchmarking \u003cem\u003ein silico\u003c/em\u003e data\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"7\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eWGS\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003eRNA-seq\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber spiked\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMean AF†\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNumber spiked\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMean AF†\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eΔAF‡\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSNV\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48,000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.05 ~ 0.5\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e48,000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.05 ~ 0.5\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e~ 0\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFalse positive DVR\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.05 ~ 0.5\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2,000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1 ~ 0.3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.05 ~ 0.1\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRV\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e~ 44,000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1 ~ 0.3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e~ 0\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDVR\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e~ 6000\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1 ~ 0.3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.05 ~ 0.1\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"7\"\u003e†: AF: \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eA\u003c/span\u003elternative allele \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eF\u003c/span\u003erequency\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"7\"\u003e‡: ΔAF = AF\u003csub\u003etreatment\u003c/sub\u003e - AF\u003csub\u003econtrol\u003c/sub\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eThe benchmarking extended to evaluating the impact of the 'boost recalibration' procedure implemented within the CADRES pipeline, compared to its omission. Concurrently, the same simulated datasets were analysed using JACUSA2, employing both RRD and RDD modes. To maximize the potential of DVR detection, a combined analysis of JACUSA2 was also conducted, where only variants identified by both RRD and RDD methods were retained, aligning with our stringent criteria for defining a DVR. Furthermore, the performances of VaDiR and rMATS-DVR methodologies were also evaluated. A comparative analysis of these methods is detailed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, providing a comprehensive overview of their efficacy in identifying DVRs.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of methodologies evaluated in the benchmarking study\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"9\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBQSR\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCompares RRD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCompares RDD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSupports comparison between conditions\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNumber of replicates supported\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eMutation calling engine\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eStatistical analysis on variant depth\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eAdditional mapping artefact filters\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCADRES\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGATK Mutect2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003erMATS GLMM\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCADRES w/o boost recalibration\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGATK Mutect2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003erMATS GLMM\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJACUSA2 joint RDD-RRD\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eJACUSA\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eDirichlet-multinomial distribution\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJACUSA2 RRD mode\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eJACUSA\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eDirichlet-multinomial distribution\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003erMATS-DVR\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGATK UnifiedGenotyper\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003erMATS GLMM\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJACUSA2 RDD mode\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e≥ 1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eJACUSA\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eDirichlet-multinomial distribution\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVaDiR\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGATK UnifiedGenotyper, GATK Mutect2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e√\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eOur analysis focused on how the number of replicates influenced the sensitivity and specificity of these methods, with each simulation being repeated three times. As depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, while all methods demonstrated comparable true positive rates in detecting DVRs, distinct differences in precision and accuracy emerged among the different types of DVR callers. Methods that considered both RDD and RRD, such as CADRES and JACUSA2 joint RDD-RRD, achieved precision and accuracy scores that surpassed the limits set for RRD-only methods, indicating fewer spiked false positives. Notably, CADRES outperformed the JACUSA2 joint RDD-RRD mode in terms of precision by 10–15%, albeit at a slight cost of 3–5% in recall effectiveness. The boost recalibration procedure proved beneficial, enhancing both true positive rates (TPR) and precision, highlighting its importance in the CADRES pipeline. The inclusion of replicates generally improved benchmark scores, though the marginal benefits diminished when the number of replicates exceeded four.\u003c/p\u003e\u003cp\u003eThese findings substantiate that CADRES, particularly when applied with the boost recalibration technique and an optimal number of replicates, serves as an effective DVR detection tool, offering superior precision in identifying meaningful RNA editing events.\u003c/p\u003e\u003ch2\u003e2.3. Real-World Evaluation of DVR Detection with Inducible A3B Cell Models\u003c/h2\u003e\u003cp\u003eTo robustly assess the effectiveness of the CADRES pipeline, we employed a real-world biological model using lentiviral inducible cells engineered to express the DNA deaminase A3B (recombinant A3B-GFP) upon doxycycline induction. This setup enables direct comparison of A3B-induced samples with their non-induced counterparts, facilitating the detection of A3B-mediated C \u0026gt; U RNA editing. A3B was selected as a benchmark due to its well-documented role as an RNA editing enzyme and its distinct RNA editing motif (5’-UUCM, where M = C or A) \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e, different from its DNA motif (5’-TCW) \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, which aids in validating RNA editing identifications.\u003c/p\u003e\u003cp\u003eFor this study, we utilized two distinct datasets: one from T-47D inducible cells, previously reported by our group \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e, and a newly generated dataset from SK-OV-3 cells. T-47D cells exhibit moderate endogenous A3B expression \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e (\u003cb\u003eFigure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e), providing a suitable background for detecting A3B-mediated DVRs, enhanced by the presence of A3B interactors. In contrast, SK-OV-3 cells, which show negligible endogenous A3B expression (\u003cb\u003eFigure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e), serve as a cleaner model to confirm DVRs specific to ectopically-expressed A3B. The concentration of doxycycline was meticulously controlled to induce A3B-GFP expression at levels 15–20 times higher than endogenous A3B in T-47D cells, while in SK-OV-3 cells, ectopic A3B-GFP levels were comparable to the endogenous A3B in T-47D cells.\u003c/p\u003e\u003cp\u003eWGS was performed at over 30× coverage, and RNA-seq was conducted with four replicates for both induced and non-induced conditions. We evaluated the performance of CADRES alongside other methods including JACUSA2 joint RDD-RRD, rMATS-DVR, and VaDiR, across these datasets. As illustrated in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, all methods detected a higher number of A \u0026gt; G(I) and C \u0026gt; U variants compared to other variant types, suggesting successful identification of RNA editing events in T-47D cells. To confirm the biological relevance of identified DVRs, we analysed the minimum folding energy (MFE) of the sequences flanking DVRs using the RNAFold algorithm \u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. The MFE profiles of DVRs identified by all methodologies closely resembled those of established RNA editing sites, rather than SNVs or A3B-mediated DNA editing sites identified in a previous study\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e (GSE193225) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). This similarity indicates that the DVRs are indeed localized within RNA transcripts and are structured in a manner conducive to RNA folding. Further analysis of the nucleotide sequences flanking these editing sites showed that CADRES and JACUSA2 joint RDD-RRD methods performed best in enriching the ‘5’-UUCM’ motif for C \u0026gt; U DVRs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb, S2), known to be preferentially targeted by A3B \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. This motif enrichment was notably weaker in results from JACUSA2 RDD or RRD methods (\u003cb\u003eFigure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/b\u003e), potentially due to the inclusion of a few non-A3B-mediated DVRs or SNVs.\u003c/p\u003e\u003cp\u003eA noticeable overlap was observed between DVRs identified by rMATS-DVR and SNVs present in T-47D cells (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). This overlap is expected as rMATS-DVR does not incorporate a mechanism to filter out SNVs. Given that rMATS-DVR and CADRES utilise similar analytical methodologies, we explored whether the inclusion of SNV contaminants might have influenced the outcomes of the rMATS GLMM model. As demonstrated in \u003cb\u003eFigure \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e\u003c/b\u003e, the inclusion of SNVs in the rMATS-DVR analysis resulted in a significant proportion (33%) of C \u0026gt; U DVRs showing a decrease in alternative allele frequency following A3B-GFP induction, a pattern not observed in CADRES-derived DVRs. Typically, A3B induction is associated with an increase in alternative allele frequency at C \u0026gt; U DVRs, suggesting that the presence of SNVs may disrupt this expected pattern.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of number of DVRs identified by various methodologies in T-47D cells induced with A3B-GFP.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"7\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCADRES\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eJACUSA2 RDD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eJACUSA2 RRD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eJACUSA2 joint RDD-RRD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eVaDiR\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003erMATs-DVR\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC \u0026gt; U DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e816\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3399\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1422\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1048\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1114\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1540\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eA \u0026gt; G(I) DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3764\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e82481\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e17753\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e14920\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7485\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e26255\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther types of DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e29\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e11367\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1791\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e396\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e8876\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3230\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC \u0026gt; U DVRs overlapping SNVs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e878\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eA \u0026gt; G(I) DVRs overlapping SNVs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3150\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther types of DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1099\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eIn a further test using the SK-OV-3 cell model, which expresses ectopic A3B at a lower but biological relevant level, CADRES demonstrated a slightly reduced sensitivity in DVRS compared to JACUSA2 joint RDD-RRD (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea\u003cb\u003e)\u003c/b\u003e but achieved better enrichment for the specific editing motifs of A \u0026gt; G(I) and C \u0026gt; U. This finding underscores that while CADRES sacrifices some sensitivity to enhance specificity, it remains effective at detecting DVRs from biologically relevant alterations (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb).\u003c/p\u003e\u003cp\u003eCollectively, these results affirm that the CADRES pipeline, when applied in conjunction with appropriate biological models and rigorous analytical techniques, provides highly specific and reliable detection of A3B-mediated DVRs across different experimental setups.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of number of DVRs identified by various methodologies in SK-OV-3 cells induced with A3B-GFP.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCADRES\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eJACUSA2 RDD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eJACUSA2 RRD\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eJACUSA2 joint RDD-RRD\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC \u0026gt; U DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e442\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2539\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e758\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e603\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eA \u0026gt; G(I) DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e165\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e63607\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11293\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e9386\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther types of DVRs\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e29375\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2179\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1199\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003ch2\u003e2.4. Validation of DVRs\u003c/h2\u003e\u003cp\u003eTo further validate the DVRs identified by CADRES, we investigated the A3B binding at regions flanking these DVRs. Utilizing previously conducted \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ee\u003c/span\u003enhanced \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eC\u003c/span\u003eross-\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003el\u003c/span\u003einking and \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eI\u003c/span\u003emmuno\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ep\u003c/span\u003erecipitation sequencing (eCLIP-seq) experiments that mapped A3B binding sites in RNA \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e (curated in GSE193225), we analysed the enrichment of A3B binding signals around the DVRs detected by CADRES, JACUSA2 RDD-RRD, rMATS-DVR and VaDiR methods. Consistent with expectations for a method demonstrating high accuracy, we observed significant enrichment of A3B binding signals at C \u0026gt; U DVRs derived from CADRES (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). This binding pattern, particularly evident immediately upstream to the C \u0026gt; U DVRs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb), confirms the presence of the A3B complex, which interferes with reverse transcription during library preparation. This validates that A3B specifically binds to RNA sites undergoing C \u0026gt; U editing, substantiating the successful identification of A3B-mediated DVRs. On the other hand, for JACUSA2-derived and rMATS-DVRs, enrichment of eCLIP-seq signal are only evident for high score DVRs, suggesting proportion of DVRs identified were not adjacent to A3B binding site (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). Quantitative analysis revealed that CADRES-derived DVRs showed a higher level of A3B binding signal enrichment compared to those identified by JACUSA2-RDD-RRD, with CADRES showing approximately ~ 15–30% greater enrichment. To explore how this difference could arise we compared DVRs identified by CADRES and JACUSA2 joint RDD-RRD methods. As is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, less than half of the DVRs from these methods intersect with each other, indicating significant different groups of DVRs being identified. Sequence motif analysis revealed that CADRES-derived DVRs display the distinct 5’-UUCM motif for A3B editing, while the JACUSA2-RDD-RRD-specific DVRs had a relatively less enrichment for A3B editing. This suggests that CADRES not only identifies A3B-derived C \u0026gt; U DVRs but does so with higher precision and specificity.\u003c/p\u003e"},{"header":"3. Discussion","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe CADRES pipeline developed in this study introduces a methodological enhancement for the detection of RNA editing events, particularly for DVRs. By seamlessly integrating sophisticated DNA/RNA variant calling with comprehensive statistical analyses, this tool refines the accuracy and specificity of RNA editing detection. The effective method for detection of DVRs not only provides a useful tool for probing the biological functions of RNA editing enzymes, but also facilitates detailed studies on how RNA editing is influenced by various experimental conditions, diseases, and therapeutic interventions. This capability allows researchers to dissect the complex mechanisms by which RNA editing impacts gene expression regulation, offering insights into disease progression and response to treatment.\u003c/p\u003e \u003cp\u003eA pivotal attribute of the CADRES pipeline is its integration of both DNA-RNA and RNA-RNA comparative analyses. This dual comparison strategy effectively mitigates confounding factors from SNVs, which can introduce apparent RNA sequence variations during transcription. Additionally, the utilisation of the rMATS GLMM statistical model enables precise detection of differential RNA editing events \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Although the rigorous criteria employed by CADRES may result in a modest reduction in sensitivity compared to other methodologies, this is counterbalanced by a significant increase in precision. This high level of precision is critical in contexts where exact identification of DVRs is paramount, such as in functional studies of RNA editing enzymes or in clinical applications where precise biomarker identification can guide therapeutic decisions. In comparison to widely-used tools like JACUSA2, CADRES offers an enhanced capability for situations where precision is prioritized over sensitivity, providing researchers and clinicians with a robust tool for the accurate dissection of RNA editing dynamics.\u003c/p\u003e \u003cp\u003eAn essential enhancement in the CADRES pipeline is the tailored application of base BQSR for sequencing reads \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. It is widely acknowledged among genomic researchers that BQSR is critical for all read alignments before mutation calling to address the common issue of pervasive sequence artifacts \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Typically, mismatches are considered artifacts by default, leading to their downgraded quality scores during recalibration, unless they coincide with pre-identified mutations which are less affected by the recalibration. While this strategy effectively isolates high-confidence SNVs, it can inadvertently diminish the detection sensitivity for RNA variants due to the incomplete cataloguing of known RNA editing sites. To mitigate this, the CADRES pipeline includes a 'boost recalibration' stage, involving a preliminary round of joint DNA-RNA mutation calling via GATK4 Mutect2. This process generates a list of tentative RVs that informs the subsequent BQSR of RNA-seq data. Our evaluations using both simulated and real datasets indicate that this modification slightly reduces the sensitivity loss typically associated with BQSR\u0026mdash;less than 8% compared to methods that omit BQSR. This approach not only retains the method's precision by minimizing sequence artifacts but also ensures a balanced detection capability, crucial for accurate RNA editing analysis.\u003c/p\u003e \u003cp\u003eThe development of high-throughput sequencing technologies has substantially enhanced our capacity to study RNA editing. While the detection of A\u0026thinsp;\u0026gt;\u0026thinsp;I editing is relatively well-established due to its prevalence and biochemical detectability, the identification of C\u0026thinsp;\u0026gt;\u0026thinsp;U editing events remains challenging \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. This is evidenced by the comparatively small repository of validated C\u0026thinsp;\u0026gt;\u0026thinsp;U editing sites, highlighting the inherent difficulties associated with accurately calling these modifications. In this study, we utilised an inducible APOBEC3B deaminase model to benchmark the performance of the CADRES pipeline in identifying C\u0026thinsp;\u0026gt;\u0026thinsp;U edits with high precision. Our findings suggest that CADRES offers an improvement in the detection of C\u0026thinsp;\u0026gt;\u0026thinsp;U editing events over existing methodologies. This capability underscores the importance of revisiting existing RNA-seq datasets, particularly those paired with whole-genome or whole-exome sequencing, to re-analyse them under more stringent criteria provided by CADRES. Looking forward, it would also be prudent to employ CADRES for the exploration of other less common types of RNA editing, such as U\u0026thinsp;\u0026gt;\u0026thinsp;C \u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e, where traditional detection methods often fall short due to confounding factors like sequencing errors and SNVs.\u003c/p\u003e \u003cp\u003eThe CADRES methodology, while robust, presents avenues for further refinement to enhance its precision and adaptability, especially in complex biological samples. Our analysis indicates that despite employing WGS to filter out SNVs, residual shallow mutations continue to influence the accuracy of downstream analyses. This issue is particularly pronounced in samples with genomic heterogeneity, such as those derived from tumour tissues. An alternative approach could involve the use of WES, which typically offers higher sequencing depth and could more effectively mitigate the impact of these mutations. However, this substitution requires careful consideration of the genomic regions not covered by WES probes, which may also hold biological relevance. Furthermore, the precision of the CADRES pipeline could be substantially improved by expanding the database of known RNA editing sites. The accuracy of the BQSR process is directly proportional to the comprehensiveness of the RNA editing database utilized. As the scientific community continues to discover and catalogue more RNA editing events \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, and as collaboration between researchers and database curators strengthens, we anticipate significant advancements in the fidelity of RNA editing detection. This evolution will enhance our ability to accurately profile RNA editing landscapes, paving the way for more nuanced biological insights and therapeutic strategies.\u003c/p\u003e \u003cp\u003eIn the development of the CADRES methodology, a deliberate trade-off was made to prioritize high predictive precision at the cost of a slight reduction in detection sensitivity. This approach utilizes stringent search criteria to ensure the accuracy of identified RNA editing sites. The question of whether this trade-off is justified merits consideration. We posit that the answer is affirmative, particularly in contexts such as RNA-editing enzyme research and the development of biomarkers, where the utility of a smaller number of highly precise RNA editing sites is evident. Furthermore, the burgeoning fields of deep learning and artificial intelligence, which rely heavily on the quality of input data for model training, stand to benefit significantly from the high precision of data provided by CADRES. While the sensitivity limitations of CADRES might be viewed as a drawback, these can potentially be mitigated by the inherent capacity of these computational methods to process large datasets efficiently. Looking ahead, we are optimistic about the integration of CADRES with advanced computational strategies, enhancing its utility and application in the rapidly evolving landscape of genomic research.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"4. Materials and Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Cell culturing\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe T-47D human breast cancer cell line and SK-OV-3 human ovarian cancer cell line were obtained from the American Type Culture Collection (ATCC). Lenti-X 293T cells, used for lentivirus production, were obtained from Clontech. T-47D cells were cultured in RPMI-1640 medium (Gibco) supplemented with 10 \u0026micro;g/mL bovine insulin (Sigma), while SK-OV-3 cells were maintained in McCoy\u0026rsquo;s 5A medium (Gibco). Both media were supplemented with 10% v/v fetal bovine serum (FBS, PAA) and 0.5% v/v penicillin/streptomycin (pen/strep, Gibco). Lenti-X 293T cells were cultured in Dulbecco's Modified Eagle Medium (DMEM, Gibco) supplemented with 10% v/v FBS and 0.5% v/v pen/strep. All cell lines were cultured at 37\u0026deg;C in a humidified atmosphere containing 5% CO2. Cell line authenticity was verified using short tandem repeat (STR) profiling, and all cell lines were regularly tested for mycoplasma contamination using PCR.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Generation of Stable Inducible Cell Lines\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eExponentially growing T-47D or SK-OV-3 cells were transfected with lentivirus in the presence of 10 \u0026micro;g/mL polybrene (Sigma) to facilitate viral entry. After a 48-hour transduction period, cells were selected with RPMI-1640 medium supplemented with 4 \u0026micro;g/mL puromycin (Gibco) to enrich for successfully transduced cells. Transduction efficiency was assessed by monitoring green fluorescent protein (GFP) expression and was further confirmed by quantitative reverse transcription PCR (qRT-PCR) after induction with doxycycline. To maintain the lentiviral cassettes in the T-47D and SK-OV-3 cell lines, cells were cultured continuously in medium containing 4 \u0026micro;g/mL puromycin.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Whole-genome and RNA sequencing\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eCells were harvested using trypsin digestion. Genomic DNA was extracted using a Qiagen Genomic-tips 500/G kit. Sequencing libraries for WGS with DNA nanoball (DNB) technology were constructed (Beijing Genomics Institute (BGI) Inc., Hong Kong). Sequencing was carried out using BGISEQ-500 sequencer with a mean sequencing coverage of greater than 30\u0026times; for each of the samples using 2\u0026times;150 bp configuration. Sequence reads from WGS was aligned to GRCh38 genome assembly using Burrow-Wheelers aligner \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTotal RNA were extracted from cells using the MagNA pure 96 platform (Roche), and RNA integrity number (RIN) was determined by BioAnalyzer 2100 (Agilent). RNA libraries were constructed and sequenced using poly-dT enrichment or ribosomal RNA depletion sample preparation methods in conjunction with DNB sequencing technology on a BGISEQ-500 instrument (BGI). Raw reads were aligned to GRCh38 genome assembly and GENCODE GRCh38.p13 annotation using STAR \u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, and then processed with SAMtools \u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. For RNA-seq in this study, four biological replicates were conducted.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Detection of DVRs using CADRES pipeline\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eData preprocessing: Preprocessing of sequencing alignments were carried out using Picard tools \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. The sequence alignment for WGS and RNA-seq, both in .bam format, were assigned with appropriate read groups. For samples sequenced with BGISEQ-500 platform, \u0026lsquo;COMPLETE\u0026rsquo; (Complete Genomics) was chosen as sequence platform (\u0026lsquo;RGPL\u0026thinsp;=\u0026thinsp;COMPLETE\u0026rsquo;). Duplicate sequence reads were then marked. For RNA-seq alignments, reads containing \u0026lsquo;N\u0026rsquo; CIGAR strings were split, while for WGS alignments, base quality score was recalibrated using recommended procedures provided by \u0026lsquo;GATK best practice workflow\u0026rsquo; introduced by the Broad Institute, where dbSNP150 database \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e resource were used as \u0026lsquo;known\u0026rsquo; sites.\u003c/p\u003e \u003cp\u003eBoost recalibration for RNA variants: This process involves joint variant calling and processing using the Mutect2 programme from GATK4 package\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. For each of the conditions, the RNA-seq data was defined as \u0026lsquo;Tumour\u0026rsquo; samples and WGS data was defines as is corresponding \u0026lsquo;Normal\u0026rsquo; control. The resultant variant calls were then selected using default Mutect2 filters, and the \u0026lsquo;PASS\u0026rsquo; variants were retained. The \u0026lsquo;PASS\u0026rsquo; variants for both conditions were then collated as putative RNA variants, and then combined with known RNA editing sites from REDIportal \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. The resulting list of variants were the used as \u0026lsquo;known\u0026rsquo; sites for the base recalibration procedure of each of the RNA-seq alignment file.\u003c/p\u003e \u003cp\u003eJoint DNA/RNA mutation calling for RNA variants: A second round of joint mutation calling were performed using Mutect2, with RNA-seq alignments as \u0026lsquo;Tumour\u0026rsquo; samples and WGS alignments as \u0026lsquo;Normal\u0026rsquo; samples, yielding a list of variants that undergoes filtering using standard Mutect2 filters. During this filtering step, maximum allowed mutations within one sequencing read was set to 4 instead of 2 to accommodate for the possible clustered nature for certain RNA editing enzymes, such as APOBEC3B which contains two deaminase domains form a homodimer. In addition to the default Mutect2 filters, two filters from the previously developed SNPiR pipeline were adopted to remove variants at the end of homopolymer runs and variants within multiple mapped sequences \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. This was achieved by scripts previously by Piskol et al., in conjunction with the pBLAT programme \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. After these filtering processes, the resulting variants are defined as RNA variants.\u003c/p\u003e \u003cp\u003eSampling of allele depth for RNA variants: The aligned sequence reads for each sample in both conditions were separated by reference genomic strand using SAM flag string. For each of the RNA variants, the depth for reference and alternative allele was counted in a strand-specific manner using SAMtools \u0026lsquo;mpileup\u0026rsquo; programme, which generates an allele depth table of RVs for each of the samples.\u003c/p\u003e \u003cp\u003erMATS statistical analysis: rMATS statistical analysis were performed using the previously developed rMATS-DVR programme\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Briefly, the allele depth table were subjected to rMATS (replicate Multivariate Analysis of Transcript Splicing) statistical analysis, which uses a generalized linear mixed model (GLMM) to simultaneously account for the RNA-seq estimation uncertainty in the alternative allele frequency as influenced by sequencing coverage in individual samples, and the variability in isoform ratios among replicates\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. This process yields a probability and false discovery rate (FDR) for a RV being a DVR. In this study, to be identified as a DVR the FDR requires less than 0.05, with minimal sequencing depth needs to be at least 10. The resulting DVR were then annotated with RefSeq.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.5 JACUSA2 joint RDD and RRD calling\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAligned files were processed using MarkDuplicates programme in Picard tools. For the processing of data, developer\u0026rsquo;s instructions were followed so that experimental design (\u003cem\u003ei.e.\u003c/em\u003e replications within condition groups) was recognised by JACUSA2 programme. Both RDD and RRD analyses were conducted independently. The outputs from these analyses were further processed using the JACUSA2helper package in R. Variant filtering was performed using the default settings of the programme. For the joint RDD-RRD analysis, only variants that met the criteria for both RDD and RRD were included in the subsequent analysis.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.6 Benchmark of DVR detection methods using simulated data\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eSimulated sequencing data were generated using the same method as reported by Piechotta \u003cem\u003eet al.\u003c/em\u003e \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e with the following exceptions. First, mutation-free WGS data was generated using ReSeq using developers\u0026rsquo; recommended setting, and RNA-seq data was generated using RESM with T-47D data as template for expression. Second, variant frequencies for SNVs were sampled from a previous study using publicly deposited data, which was applied to both WGS and RNA-seq samples. Third, two types of variants were spiked into mutation-free simulated data, \u003cem\u003ei.e.\u003c/em\u003e 50,000 SNVs and 50,000 RVs, as is summarised in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Prior to benchmark analysis on simulated data, statistical analysis on spiked variants using rMATS GLMM and/or JACUSA methods were conducted to distinguish DVRs from other RVs, as well as calculating the theoretical best for variant calling methods. The identification of DVRs were defined as true positive in the following analysis.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.7 Benchmark of DVR detection using inducible A3B deaminase models\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eFor inducible expression of the lentiviral-encoded proteins, T-47D and SK-OV-3 cells cells were exposed to culture media supplemented with 100 ng/mL doxycycline (Sigma) to trigger protein expression. After 72 hours of induction, the expression of ectopic A3B were monitored by RT-qPCR, ensuring that the ectopic A3B is 10\u0026ndash;20 fold expression to its endogenous counterpart. The cell samples were subjected to WGS and RNA-seq.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.8 DVR motif analysis\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eSequences flanking the identified DVRs were extracted using the GRChg38 reference sequence. The nucleotide composition around these sites was visualized using Weblogo3 to analyse sequence conservation and motif prevalence \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.9 RNA folding analysis\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eRNA secondary structure predictions were performed using the RNAFold programme \u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e, supported by the LncFinder package from R \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, which aids in identifying potential long non-coding RNA structures. For this analysis, a window of \u0026plusmn;\u0026thinsp;100 base pairs around each identified SNV or RNA editing site was considered. To ensure a robust comparison, SNVs were sampled from the dbSNP database \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e with a random selection of 100,000 variants. Control sequences were selected based on RNA-seq data with a read depth greater than 10, using BEDtools to isolate relevant gene segments. These segments were then randomly sampled (n\u0026thinsp;=\u0026thinsp;1,000) to match the sequence motifs of interest \u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. The folding conditions for the RNAFold analysis were set to physiological conditions (37\u0026deg;C, with a salt concentration of 1.0 M and no lonely pairs, -T 37 -salt 1.0 -d 2 --noLP), which are crucial for accurately modelling RNA folding dynamics in a cellular environment.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.10 Visualisation of eCLIP-seq data\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eeCLIP-seq data were converted to bigwig file format, which denotes fold change of immunoprecipitated signal over IgG size-matched control. The files were then visualised using Seqplots (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://przemol.github.io/seqplots/\u003c/span\u003e\u003cspan address=\"https://przemol.github.io/seqplots/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) programme.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e Conceptualisation, C.Z. and X.L.; methodology, C.Z.; software, J.S.; investigation, J.S. and C.Z.; writing, C.Z. and J.S.; visualisation, C.Z. and J.S.; supervision, X.L.; project administration, X.L.; funding acquisition, C.Z. and X.L. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u003c/strong\u003e CADRES is coded by a combination of Python and Shell script languages, which is available at GitHub (https://github.com/junsun-hash/CADRES). Sequencing data related to this study is available at GEO with accession numbers (GSE193225, GSExxxxx).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments:\u003c/strong\u003e The authors thank the funding support from Shanghai Institute of Biological Products. C.Z. was sponsored by Shanghai Pujiang Programme (22PJD104), and Science and Technology Commission of Shanghai (23S11901100).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of Interest:\u003c/strong\u003e J.S., C.Z., and X.L. are employees of Shanghai Institute of Biological Products, an entity presently engaged in the commercial development of therapeutic biologics.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eQiu, L., Jing, Q., Li, Y. \u0026amp; Han, J. RNA modification: mechanisms and therapeutic targets. Mol Biomed 4, 25 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEisenberg, E. \u0026amp; Levanon, E. Y. A-to-I RNA editing - immune protector and transcriptome diversifier. Nat Rev Genet 19, 473\u0026ndash;490 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlotkin, W. \u0026amp; Nishikura, K. Adenosine-to-inosine RNA editing and human disease. Genome Med 5, 105 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePecori, R., Di Giorgio, S., Paulo Lorenzo, J. \u0026amp; Nina Papavasiliou, F. Functions and consequences of AID/APOBEC-mediated DNA and RNA deamination. Nat Rev Genet 23, 505\u0026ndash;518 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi, H., Chai, P., Jia, R. \u0026amp; Fan, X. Novel insight into the regulatory roles of diverse RNA modifications: Re-defining the bridge between transcription and translation. Mol Cancer 19, 78 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSatam, H. \u003cem\u003eet al.\u003c/em\u003e Next-Generation Sequencing Technology: Current Trends and Advancements. Biology (Basel) 12 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalkley, C. R. \u0026amp; Li, J. B. Rewriting the transcriptome: adenosine-to-inosine RNA editing by ADARs. Genome Biol 18, 205 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, S. B. \u0026amp; Li, J. B. RNA editing and immune control: from mechanism to therapy. Curr Opin Genet Dev 86, 102195 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalter, J. D., Bennett, R. P. \u0026amp; Smith, H. C. The APOBEC Protein Family: United by Structure, Divergent in Function. Trends Biochem Sci 41, 578\u0026ndash;594 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlanc, V. \u0026amp; Davidson, N. O. APOBEC-1-mediated RNA editing. Wiley Interdiscip Rev Syst Biol Med 2, 594\u0026ndash;602 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma, S. \u003cem\u003eet al.\u003c/em\u003e APOBEC3A cytidine deaminase induces RNA editing in monocytes and macrophages. Nat Commun 6, 6881 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaysal, B. E. \u003cem\u003eet al.\u003c/em\u003e Hypoxia-inducible C-to-U coding RNA editing downregulates SDHB in monocytes. PeerJ 1, e152 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlqassim, E. Y. \u003cem\u003eet al.\u003c/em\u003e RNA editing enzyme APOBEC3A promotes pro-inflammatory M1 macrophage polarization. Commun Biol 4, 102 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan der Auwera, G. A. \u003cem\u003eet al.\u003c/em\u003e From FastQ data to high confidence variant calls: the Genome Analysis Toolkit best practices pipeline. Curr Protoc Bioinformatics 43, 11 10 11\u0026ndash;11 10 33 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eBest Practices for Variant Calling with the GATK\u003c/em\u003e, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u0026lt;https://www.broadinstitute.org/partnerships/education/broade/best-practices-variant-calling-gatk-1\u0026gt;\u003c/span\u003e\u003cspan address=\"http://%3Chttps://www.broadinstitute.org/partnerships/education/broade/best-practices-variant-calling-gatk-1%3E\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBahn, J. H. \u003cem\u003eet al.\u003c/em\u003e Accurate identification of A-to-I RNA editing in human by transcriptome sequencing. Genome Res 22, 142\u0026ndash;150 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiskol, R., Ramaswami, G. \u0026amp; Li, J. B. Reliable identification of genomic variants from RNA-seq data. Am J Hum Genet 93, 641\u0026ndash;651 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, C. \u003cem\u003eet al.\u003c/em\u003e RVboost: RNA-seq variants prioritization using a boosting method. Bioinformatics 30, 3414\u0026ndash;3416 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeums, L. \u003cem\u003eet al.\u003c/em\u003e VaDiR: an integrated approach to Variant Detection in RNA. Gigascience 7, 1\u0026ndash;13 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, J., Pan, Y., Shen, S., Lin, L. \u0026amp; Xing, Y. rMATS-DVR: rMATS discovery of differential variants in RNA. Bioinformatics 33, 2216\u0026ndash;2217 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiechotta, M., Naarmann-de Vries, I. S., Wang, Q., Altmuller, J. \u0026amp; Dieterich, C. RNA modification mapping with JACUSA2. Genome Biol 23, 115 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiechotta, M., Wyler, E., Ohler, U., Landthaler, M. \u0026amp; Dieterich, C. JACUSA: site-specific identification of RNA editing events from replicate sequencing data. BMC Bioinformatics 18, 7 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePicard Tools.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcKenna, A. \u003cem\u003eet al.\u003c/em\u003e The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 20, 1297\u0026ndash;1303 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCibulskis, K. \u003cem\u003eet al.\u003c/em\u003e Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat Biotechnol 31, 213\u0026ndash;219 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMansi, L. \u003cem\u003eet al.\u003c/em\u003e REDIportal: millions of novel A-to-I RNA editing events from thousands of RNAseq experiments. Nucleic Acids Res 49, D1012-D1019 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlonso de la Vega, A. \u003cem\u003eet al.\u003c/em\u003e Acute expression of human APOBEC3B in mice results in RNA editing and lethality. Genome Biol 24, 267 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, C. L., Y.; Wang, M.; Chen, B.; Xiong, F.; Mitsopoulos, C.; Rosanesse, O.; Clarke, P. Characterisation of APOBEC3B-Mediated RNA Editing in Breast Cancer Cells Reveals Regulatory Roles of NEAT1 and MALAT1 lncRNAs. Research Gate (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurns, M. B. \u003cem\u003eet al.\u003c/em\u003e APOBEC3B is an enzymatic source of mutation in breast cancer. Nature 494, 366\u0026ndash;370 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurns, M. B., Temiz, N. A. \u0026amp; Harris, R. S. Evidence for APOBEC3B mutagenesis in multiple human cancers. Nat Genet 45, 977\u0026ndash;983 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeriyasamy, M. \u003cem\u003eet al.\u003c/em\u003e APOBEC3B-Mediated Cytidine Deamination Is Required for Estrogen Receptor Action in Breast Cancer. Cell Rep 13, 108\u0026ndash;121 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGruber, A. R., Lorenz, R., Bernhart, S. H., Neubock, R. \u0026amp; Hofacker, I. L. The Vienna RNA websuite. Nucleic Acids Res 36, W70-74 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, C. \u003cem\u003eet al.\u003c/em\u003e R-loop editing by DNA cytosine deaminase APOBEC3B determines the activity of estrogen receptor enhancers. bioRxiv (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZverinova, S. \u0026amp; Guryev, V. Variant calling: Considerations, practices, and developments. Hum Mutat 43, 976\u0026ndash;985 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoboldt, D. C. Best practices for variant calling in clinical sequencing. Genome Med 12, 91 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLerner, T. \u003cem\u003eet al.\u003c/em\u003e C-to-U RNA Editing: From Computational Detection to Experimental Validation. Methods Mol Biol 2181, 51\u0026ndash;67 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnoop, V. C-to-U and U-to-C: RNA editing in plant organelles and beyond. J Exp Bot 74, 2273\u0026ndash;2294 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, H. \u0026amp; Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754\u0026ndash;1760 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarrow, J. \u003cem\u003eet al.\u003c/em\u003e GENCODE: the reference human genome annotation for The ENCODE Project. Genome Res 22, 1760\u0026ndash;1774 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDobin, A. \u003cem\u003eet al.\u003c/em\u003e STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15\u0026ndash;21 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanecek, P. \u003cem\u003eet al.\u003c/em\u003e Twelve years of SAMtools and BCFtools. Gigascience 10 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmigielski, E. M., Sirotkin, K., Ward, M. \u0026amp; Sherry, S. T. dbSNP: a database of single nucleotide polymorphisms. Nucleic Acids Res 28, 352\u0026ndash;355 (2000).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSherry, S. T., Ward, M. \u0026amp; Sirotkin, K. dbSNP-database for single nucleotide polymorphisms and other classes of minor genetic variation. Genome Res 9, 677\u0026ndash;679 (1999).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, M. \u0026amp; Kong, L. pblat: a multithread blat algorithm speeding up aligning sequences to genomes. BMC Bioinformatics 20, 28 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen, S. \u003cem\u003eet al.\u003c/em\u003e rMATS: robust and flexible detection of differential alternative splicing from replicate RNA-Seq data. Proc Natl Acad Sci U S A 111, E5593-5601 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrooks, G. E., Hon, G., Chandonia, J. M. \u0026amp; Brenner, S. E. WebLogo: a sequence logo generator. Genome Res 14, 1188\u0026ndash;1190 (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan, S. \u003cem\u003eet al.\u003c/em\u003e LncFinder: an integrated platform for long non-coding RNA identification utilizing sequence intrinsic composition, structural information and physicochemical property. Brief Bioinform 20, 2009\u0026ndash;2027 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuinlan, A. R. \u0026amp; Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841\u0026ndash;842 (2010).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"RNA editing, Next-generation sequencing, Differential Variants on RNA (DVRs), Bioinformatics, Variant calling, APOBEC deaminase","lastPublishedDoi":"10.21203/rs.3.rs-4855669/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4855669/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRNA editing is a post-transcriptional modification critical for gene regulation and adaptation. Detecting these modifications, especially \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifferential \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eV\u003c/span\u003eariants on \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA (DVRs), presents significant challenges due to the subtlety of RNA editing sites and the complexity of RNA sequences. To improve the detection and analysis of RNA editing sites, we developed the \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eCa\u003c/span\u003elibrated \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eD\u003c/span\u003eifferential \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003eNA \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eE\u003c/span\u003editing \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eS\u003c/span\u003ecanner (CADRES), an analytical pipeline that combines sophisticated DNA/RNA variant calling with detailed statistical analysis. This study validates CADRES through rigorous \u003cem\u003ein silico\u003c/em\u003e and experimental dataset using inducible cell models of the APOBEC3B (A3B) deaminase. CADRES demonstrates improved specificity and accuracy over existing methodologies, effectively identifying A3B-mediated C\u0026thinsp;\u0026gt;\u0026thinsp;U edits and distinguishing these from related sequencing artifacts and DNA polymorphisms. The adaptability of CADRES is highlighted by its consistent performance across varied experimental conditions and different numbers of replicates. Our findings illustrate that CADRES provides a reliable tool for the precise identification of RNA editing sites, contributing valuable insights into RNA editing dynamics. This capability is crucial for advancing our understanding of the molecular mechanisms underlying gene expression regulation and the potential development of therapeutic strategies.\u003c/p\u003e","manuscriptTitle":"CADRES: An Analytical Pipeline for Precise Identification of Differential RNA Editing Events Across Varied Biological Conditions","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-28 05:30:27","doi":"10.21203/rs.3.rs-4855669/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-01-10T09:32:45+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-16T09:00:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"239888054573335508786268665573663474238","date":"2024-12-02T06:23:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"104334200840285858142891345660242519381","date":"2024-10-29T21:31:13+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-09-27T08:39:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"156826486318324271318802265443109122458","date":"2024-09-19T01:18:49+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-09-02T16:30:11+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-08-24T21:04:40+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-08-20T05:10:13+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-08-14T14:09:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-08-04T06:45:31+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3e06d0b6-ef8a-41fd-b2ac-7ca375695d08","owner":[],"postedDate":"April 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":35550919,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":35550920,"name":"Biological sciences/Molecular biology/Rna metabolism/Rna editing"},{"id":35550921,"name":"Biological sciences/Molecular biology/Rna metabolism/Rna modification"},{"id":35550922,"name":"Biological sciences/Biotechnology/Sequencing/Dna sequencing"},{"id":35550923,"name":"Biological sciences/Biotechnology/Sequencing/Next generation sequencing"},{"id":35550924,"name":"Biological sciences/Biotechnology/Sequencing/Rna sequencing"}],"tags":[],"updatedAt":"2025-06-09T16:09:15+00:00","versionOfRecord":{"articleIdentity":"rs-4855669","link":"https://doi.org/10.1038/s41598-025-04957-7","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-06-04 15:57:31","publishedOnDateReadable":"June 4th, 2025"},"versionCreatedAt":"2025-04-28 05:30:27","video":"","vorDoi":"10.1038/s41598-025-04957-7","vorDoiUrl":"https://doi.org/10.1038/s41598-025-04957-7","workflowStages":[]},"version":"v1","identity":"rs-4855669","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4855669","identity":"rs-4855669","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}