Calling small variants using universality with Bayes-factor-adjusted odds ratios

Abstract

The application of next-generation sequencing in research and particularly in clinical routine requires highly accurate variant calling. Here we describe UVC, a method for calling small variants of germline or somatic origin. By unifying opposite assumptions with sublation, we discovered the following two empirical laws to improve variant calling: allele fraction at high sequencing depth is inversely proportional to the cubic root of variant-calling error rate, and odds ratios adjusted with Bayes factors can model various sequencing biases. UVC outperformed other variant callers on the GIAB germline truth sets, 192 scenarios of in silico mixtures simulating 192 combinations of tumor/normal sequencing depths and tumor/normal purities, the GIAB somatic truth sets derived from physical mixture, and the SEQC2 somatic reference sets derived from the breast-cancer cell-line HCC1395. UVC achieved 100% concordance with the manual review conducted by multiple independent researchers on a Qiagen 71-gene-panel dataset derived from 16 patients with colon adenoma. UVC outperformed other unique molecular identifier (UMI)-aware variant callers on the datasets used for publishing these variant callers. Performance was measured with sensitivity-specificity trade off for called variants. The improved variant calls generated by UVC from previously published UMI-based sequencing data provided additional insight about DNA damage repair. UVC is open-sourced under the BSD 3-Clause license at https://github.com/genetronhealth/uvc and quay.io/genetronhealth/gcc-6-3-0-uvc-0-6-0-441a694

NeoHeadHunter: an algorithm for the detection, ranking and probabilistic classification of neoepitope candidates

“…Let DtnQUAL( x ) be the allele quality of the mutation that is transcribed and translated into the peptide of x , where the allele quality is defined by the QUAL column in the VCF file format specification [Danecek et al, 2011] and computed by UVC in tumor-normal-paired mode from DNA-seq data [Zhao et al, 2022]. Let DtAD( x ), DtAF( x ) and DnAF( x ) be the tumor allele depth, tumor allele fraction and normal allele fraction (of the mutation transcribed and translated into the peptide of x ) computed by UVC from DNA-seq data, respectively [Zhao et al, 2022]. Let RtAD( x ) and RtAF( x ) be the tumor mutant allele depth and tumor mutant allele fraction (of the mutation translated into the peptide of x ) computed by UVC from RNA-seq data, respectively [Zhao et al, 2022].…”

“…Let DtAD( x ), DtAF( x ) and DnAF( x ) be the tumor allele depth, tumor allele fraction and normal allele fraction (of the mutation transcribed and translated into the peptide of x ) computed by UVC from DNA-seq data, respectively [Zhao et al, 2022]. Let RtAD( x ) and RtAF( x ) be the tumor mutant allele depth and tumor mutant allele fraction (of the mutation translated into the peptide of x ) computed by UVC from RNA-seq data, respectively [Zhao et al, 2022]. Let Affinity( x ) and BindLevel( x ) be the binding affinity (with the unit of nanomolar of dissociation constant) and binding level of the pMHC x computed by netMHCpan, respectively [Reynisson et al, 2020].…”

“…2). We used BWA MEM version 0.7.17 [Li, 2013] for DNA-seq alignments, STAR version 2.7.8 [Dobin et al, 2013] for RNA-seq alignments, STAR-Fusion version 1.12.0 [Haas et al, 2017] for RNA fusion detection, ASNEO commit 9f43cff [Zhang et al, 2020] for RNA splicing-variant detection, Kallisto version 0.48.0 [Bray et al, 2016] for RNA-transcript abundance estimation, UVC version 0.14.2.15f4adc and uvc-delins version 0.1.7.30af093 [Zhao et al, 2022] for small-variant calling, Ensembl-VEP version 109.3 [McLaren et al, 2016] for small-variant effect prediction, Optitype version 1.3.2 [Szolek et al, 2014] with sequencing data pre-filtered by bwa mem for HLA typing, NetMHCpan version 4.1b [Reynisson et al, 2020] for peptide-MHC binding affinity prediction and NetMHCstabpan version 1.0 along with NetMHCpan version 2.8 [Rasmussen et al, 2016] for peptide-MHC binding stability prediction. Additionally, we used the CTAT resource library [Haas et al, 2019] for DNA-seq alignments, RNA-seq alignments and fusion detection.…”

SafeMut: UMI-aware variant simulator incorporating allele-fraction overdispersion in read editing

“…Second, allele ratio (the number of fragments A supporting the ALT mutant variant allele, divided by the number of fragments B supporting all non-ALT alleles) is known to follow a log-normal distribution [19]. Additionally, the event A/B = 1 is known to be 10 (15/10) times more likely than the event A/B = 2 [19]. Hence, we used such lognormal distribution to simulate the overdispersion in the fluctuation of variant allele fractions (VAFs).…”

“…In sum, our algorithm introduces some multiplicative stochastic error into F to obtain F’ such that the allele ratio follows a log-normal distribution, where the event that R 1 / R 2 = 1 is 10( 15/10) times as likely as the event that R 1 / R 2 = 2 given two allele ratios R 1 and R 2 . This log-normal distribution has been discovered and used in the literature [19, 22].…”