Variant Calling Tools Research Articles

Best practices for benchmarking germline small-variant calls in human genomes.

Standardized benchmarking approaches are required to assess the accuracy of variants called from sequence data. Although variant-calling tools and the metrics used to assess their performance continue to improve, important challenges remain. Here, as part of the Global Alliance for Genomics and Health (GA4GH), we present a benchmarking framework for variant calling. We provide guidance on how to match variant calls with different representations, define standard performance metrics, and stratify performance by variant type and genome context. We describe limitations of high-confidence calls and regions that can be used as truth sets (for example, single-nucleotide variant concordance of two methods is 99.7% inside versus 76.5% outside high-confidence regions). Our web-based app enables comparison of variant calls against truth sets to obtain a standardized performance report. Our approach has been piloted in the PrecisionFDA variant-calling challenges to identify the best-in-class variant-calling methods within high-confidence regions. Finally, we recommend a set of best practices for using our tools and evaluating the results.

Comparison of three variant callers for human whole genome sequencing

Testing of patients with genetics-related disorders is in progress of shifting from single gene assays to gene panel sequencing, whole-exome sequencing (WES) and whole-genome sequencing (WGS). Since WGS is unquestionably becoming a new foundation for molecular analyses, we decided to compare three currently used tools for variant calling of human whole genome sequencing data. We tested DeepVariant, a new TensorFlow machine learning-based variant caller, and compared this tool to GATK 4.0 and SpeedSeq, using 30×, 15× and 10× WGS data of the well-known NA12878 DNA reference sample. According to our comparison, the performance on SNV calling was almost similar in 30× data, with all three variant callers reaching F-Scores (i.e. harmonic mean of recall and precision) equal to 0.98. In contrast, DeepVariant was more precise in indel calling than GATK and SpeedSeq, as demonstrated by F-Scores of 0.94, 0.90 and 0.84, respectively. We conclude that the DeepVariant tool has great potential and usefulness for analysis of WGS data in medical genetics.

Development of a low-frequency variant calling tool for the analysis of tumor samples processed with ultra-deep next-generation sequencing

Development of a low-frequency variant calling tool for the analysis of tumor samples processed with ultra-deep next-generation sequencing

Evaluating the performance of tools used to call minority variants from whole genome short-read data.

Background: High-throughput whole genome sequencing facilitates investigation of minority virus sub-populations from virus positive samples. Minority variants are useful in understanding within and between host diversity, population dynamics and can potentially assist in elucidating person-person transmission pathways. Several minority variant callers have been developed to describe low frequency sub-populations from whole genome sequence data. These callers differ based on bioinformatics and statistical methods used to discriminate sequencing errors from low-frequency variants. Methods: We evaluated the diagnostic performance and concordance between published minority variant callers used in identifying minority variants from whole-genome sequence data from virus samples. We used the ART-Illumina read simulation tool to generate three artificial short-read datasets of varying coverage and error profiles from an RSV reference genome. The datasets were spiked with nucleotide variants at predetermined positions and frequencies. Variants were called using FreeBayes, LoFreq, Vardict, and VarScan2. The variant callers' agreement in identifying known variants was quantified using two measures; concordance accuracy and the inter-caller concordance. Results: The variant callers reported differences in identifying minority variants from the datasets. Concordance accuracy and inter-caller concordance were positively correlated with sample coverage. FreeBayes identified the majority of variants although it was characterised by variable sensitivity and precision in addition to a high false positive rate relative to the other minority variant callers and which varied with sample coverage. LoFreq was the most conservative caller. Conclusions: We conducted a performance and concordance evaluation of four minority variant calling tools used to identify and quantify low frequency variants. Inconsistency in the quality of sequenced samples impacts on sensitivity and accuracy of minority variant callers. Our study suggests that combining at least three tools when identifying minority variants is useful infiltering errors when calling low frequency variants.

Evaluating the performance of tools used to call minority variants from whole genome short-read data

Background: High-throughput whole genome sequencing facilitates investigation of minority virus sub-populations from virus positive samples. Minority variants are useful in understanding within and between host diversity, population dynamics and can potentially assist in elucidating person-person transmission pathways. Several minority variant callers have been developed to describe low frequency sub-populations from whole genome sequence data. These callers differ based on bioinformatics and statistical methods used to discriminate sequencing errors from low-frequency variants. Methods: We evaluated the diagnostic performance and concordance between published minority variant callers used in identifying minority variants from whole-genome sequence data from virus samples. We used the ART-Illumina read simulation tool to generate three artificial short-read datasets of varying coverage and error profiles from an RSV reference genome. The datasets were spiked with nucleotide variants at predetermined positions and frequencies. Variants were called using FreeBayes, LoFreq, Vardict, and VarScan2. The variant callers’ agreement in identifying known variants was quantified using two measures; concordance accuracy and the inter-caller concordance. Results: The variant callers reported differences in identifying minority variants from the datasets. Concordance accuracy and inter-caller concordance were positively correlated with sample coverage. FreeBayes identified the majority of variants although it was characterised by variable sensitivity and precision in addition to a high false positive rate relative to the other minority variant callers and which varied with sample coverage. LoFreq was the most conservative caller. Conclusions: We conducted a performance and concordance evaluation of four minority variant calling tools used to identify and quantify low frequency variants. Inconsistency in the quality of sequenced samples impacts on sensitivity and accuracy of minority variant callers. Our study suggests that combining at least three tools when identifying minority variants is useful in filtering errors when calling low frequency variants.

Abstract 5296: R2D2: An integrated analysis framework to infer the functional impact of single nucleotide variants (SNVs) using matched germline and tumor DNA and RNA sequencing data

Abstract Introduction: Single nucleotide variants (SNVs) are the most abundant genetic variation in the human genome. Unlike protein-truncating alterations, the functional impact of SNVs can be very difficult to infer from DNA sequencing alone. Although functional techniques such as RNA-Seq have been used to quantify protein expression in normal and tumor samples, up to date, there has not been a robust tool to systematically evaluate the transcriptional perturbations caused by SNVs in the germline and somatic spaces. Here, we describe R2D2, a computational algorithm that jointly analyzes matched germline and tumor DNA and RNA sequencing data to infer the functional impact of coding SNVs. Methods: R2D2 was developed to evaluate the allelic fraction of called SNVs across matched normal and tumor genetic and transcriptomic data to classify these variants into predefined categories of potential functional and biological significance. Ideally, four sets of matched DNA and RNA variants from both normal and tumor tissues would be available to classify both germline and somatic SNVs. However, R2D2 is capable of analyzing various combinations of DNA and RNA sequencing data to infer the functional impact of detected SNVs. R2D2 has also been optimized to work with various commonly used variant calling and annotation tools such as the Genome Analysis Tool Kit (GATK) and Oncotator. Results: We used R2D2 to evaluate the functional impact of germline and somatic SNVs of matched germline and tumor whole exome and transcriptome sequencing data of five patients with primary lung adenocarcinoma. On average, standard variant calling algorithms detected 34,227 (range: 33541-35698) SNVs across all sample types of each patient in our cohort. These variants were just classified as germline and somatic by these pipelines. However, surveying the allelic fraction (AF) of each variant across all sample types using R2D2 detected unexpected patterns of AF deviation, such as imbalanced allelic expression, somatic biallelic inactivation, and aberrant RNA splicing. For example, R2D2 identified somatic loss of heterozygosity of an average of 186 (range: 70-409) SNVs per patient, which can be important when studying potential mechanisms of tumorigenesis. In addition, 13.2% (range: 9.0-17.3) of all detected SNVs had an imbalanced RNA expression where one allele was more selectively expressed in the normal or tumor tissues. Lastly, we also identified a total of 1454 (range: 1349-1546) SNVs per patient that were only detected in the normal or tumor RNA-Seq data. Conclusion: We developed R2D2 to help pinpointing functional genetic SNVs that are of potential clinical significance. R2D2 is an easy-to-implement and flexible module that can be integrated with various variant-calling pipelines to identify unexpected patterns of variant expression. Citation Format: Alma Imamovic, Saud H. AlDubayan, Nathanael Moore, Celine G. Han, Brendan Reardon, Eliezer M. Van Allen. R2D2: An integrated analysis framework to infer the functional impact of single nucleotide variants (SNVs) using matched germline and tumor DNA and RNA sequencing data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5296.

Abstract 419: Reproducibility assessment of mutations calls in exome- and whole-genome sequencing using consensus calling from TCGA and ICGC

Abstract Two large cancer genomic consortia recently published the largest and highest-quality consensus mutations calls for both whole-exome sequencing (WES) and whole-genome sequencing (WGS) in cancer: The Cancer Genome Atlas (TCGA), and the International Cancer Genetics Consortia (ICGC), respectively. Together these datasets encompass more than 60M mutations from ~13,000 samples (~10,000 WES and ~3,000 WGS). An intersecting set of 742 samples, from 22 cancer types, was sequenced using both platforms and mutations were identified using a combined 13 variant calling tools (7 WES and 5 WGS). These samples represent an ideal dataset to compare and contrast WES with WGS performance, reliability, and reproducibility of mutation calling in exons, and provide the community with key regions flanking exons that play a role in carcinogenesis. MAF files were collected using strict filtering criteria for initial file release, including the elimination of germline contaminants, 8-oxo-guanine artifacts, depth filtering and repeat masking. Additional filtering included minimum coverage requirements and restriction of both WES and WGS to variants detected within targeted exons. Finally, we restricted our data to known cancer genes. This final step suggests that these 742 samples have anywhere between 11.5K to 12.3K mutations from covered exons in potential cancer driver genes—WES and WGS, respectively. Preliminary results found that ~70% of samples had had >80% congruent mutations between both platforms; ~25% of samples had had >80% congruent mutations calls in one or the other platform; and the remaining samples had poor performance in replicating identical mutations. We observed that a majority of the variants unique to a sequencing platform were primarily from mutations with low VAF. We also sought to explore regions of the genome that are captured by both technologies despite the knowledge that WES did not target these regions. This is made possible by obtaining access to the primary data resources, and relaxing filtering criteria to include other regions such as 3' and 5' UTR, exon flanking regions, and intronic regions. We identified many recurrent mutations from non-exonic regions that were corroborated using both platforms that have not been previously reported in pan-cancer efforts. At this historic junction in time, as preliminary results from whole-genome sequencing efforts emerge and large exome sequencing efforts taper, 742 samples spanning both efforts can provide insights into the lessons learned from exome sequencing, and provide a solid foundation stepping forward into whole-genome analysis. We will continue to glean insights into the etiology of human disease by using both technologies; however, these mutation calls highlight the challenges that still exist in somatic variant calling, and provide grounds for more critical evaluation of genomic findings in cancer. Citation Format: Matthew H. Bailey, Liang-Bo Wang, Wen-Wei Liang, Steven Foltz, Guanlan Dong, Michael C. Wendl, Michael McLellan, Angela C. Hirbe, Jared Simpson, Mark Gerstein, Li Ding. Reproducibility assessment of mutations calls in exome- and whole-genome sequencing using consensus calling from TCGA and ICGC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 419.

Appreci8: a pipeline for precise variant calling integrating 8 tools.

MotivationThe application of next-generation sequencing in research and particularly in clinical routine requires valid variant calling results. However, evaluation of several commonly used tools has pointed out that not a single tool meets this requirement. False positive as well as false negative calls necessitate additional experiments and extensive manual work. Intelligent combination and output filtration of different tools could significantly improve the current situation.ResultsWe developed appreci8, an automatic variant calling pipeline for calling single nucleotide variants and short indels by combining and filtering the output of eight open-source variant calling tools, based on a novel artifact- and polymorphism score. Appreci8 was trained on two data sets from patients with myelodysplastic syndrome, covering 165 Illumina samples. Subsequently, appreci8’s performance was tested on five independent data sets, covering 513 samples. Variation in sequencing platform, target region and disease entity was considered. All calls were validated by re-sequencing on the same platform, a different platform or expert-based review. Sensitivity of appreci8 ranged between 0.93 and 1.00, while positive predictive value ranged between 0.65 and 1.00. In all cases, appreci8 showed superior performance compared to any evaluated alternative approach.Availability and implementationAppreci8 is freely available at https://hub.docker.com/r/wwuimi/appreci8/. Sequencing data (BAM files) of the 678 patients analyzed with appreci8 have been deposited into the NCBI Sequence Read Archive (BioProjectID: 388411; https://www.ncbi.nlm.nih.gov/bioproject/PRJNA388411).Supplementary information Supplementary data are available at Bioinformatics online.

A framework for the estimation of the proportion of true discoveries in single nucleotide variant detection studies for human data.

Any single nucleotide variant detection study could benefit from a fast and cheap method of measuring the quality of variant call list. It is advantageous to be able to see how the call list quality is affected by different variant filtering thresholds and other adjustments to the study parameters. Here we look into a possibility of estimating the proportion of true positives in a single nucleotide variant call list for human data. Using whole-exome and whole-genome gold standard data sets for training, we focus on building a generic model that only relies on information available from any variant caller. We assess and compare the performance of different candidate models based on their practical accuracy. We find that the generic model delivers decent accuracy most of the time. Further, we conclude that its performance could be improved substantially by leveraging the variant quality metrics that are specific to each variant calling tool.

Evaluating the performance of tools used to call minority variants from whole genome short-read data.

Background: High-throughput whole genome sequencing facilitates investigation of minority sub-populations from virus positive samples. Minority variants are useful in understanding within and between host diversity, population dynamics and can potentially help to elucidate person-person transmission chains. Several minority variant callers have been developed to describe the minority variants sub-populations from whole genome sequence data. However, they differ on bioinformatics and statistical approaches used to discriminate sequencing errors from low-frequency variants. Methods: We evaluated the diagnostic performance and concordance between published minority variant callers used in identifying minority variants from whole-genome sequence data. The ART-Illumina read simulation tool was used to generate three artificial short-read datasets of varying coverage and error profiles from an RSV reference genome. The datasets were spiked with nucleotide variants at predetermined positions and frequencies. Variants were called using FreeBayes, LoFreq, Vardict, and VarScan2. The variant callers' agreement in identifying known variants was quantified using two measures; concordance accuracy and the inter-caller concordance. Results: The variant callers reported differences in identifying minority variants from the datasets. Concordance accuracy and inter-caller concordance were positively correlated with sample coverage. FreeBayes identified majority of the variants although it was characterised by variable sensitivity and precision in addition to a high false positive rate relative to the other minority variant callers and which varied with sample coverage. LoFreq was the most conservative caller. Conclusions: We conducted a performance and concordance evaluation of four minority variant calling tools used to identify and quantify low frequency variants. Inconsistency in the quality of sequenced samples impact on sensitivity and accuracy of minority variant callers. Our study suggests that combining at least three tools when identifying minority variants is useful infiltering errors when calling low frequency variants.

CoVaCS: a consensus variant calling system

BackgroundThe advent and ongoing development of next generation sequencing technologies (NGS) has led to a rapid increase in the rate of human genome re-sequencing data, paving the way for personalized genomics and precision medicine. The body of genome resequencing data is progressively increasing underlining the need for accurate and time-effective bioinformatics systems for genotyping - a crucial prerequisite for identification of candidate causal mutations in diagnostic screens.ResultsHere we present CoVaCS, a fully automated, highly accurate system with a web based graphical interface for genotyping and variant annotation. Extensive tests on a gold standard benchmark data-set -the NA12878 Illumina platinum genome- confirm that call-sets based on our consensus strategy are completely in line with those attained by similar command line based approaches, and far more accurate than call-sets from any individual tool. Importantly our system exhibits better sensitivity and higher specificity than equivalent commercial software.ConclusionsCoVaCS offers optimized pipelines integrating state of the art tools for variant calling and annotation for whole genome sequencing (WGS), whole-exome sequencing (WES) and target-gene sequencing (TGS) data. The system is currently hosted at Cineca, and offers the speed of a HPC computing facility, a crucial consideration when large numbers of samples must be analysed. Importantly, all the analyses are performed automatically allowing high reproducibility of the results. As such, we believe that CoVaCS can be a valuable tool for the analysis of human genome resequencing studies. CoVaCS is available at: https://bioinformatics.cineca.it/covacs.

Versatile Identification of Copy Number Variants with Canvas.

Versatile and efficient variant calling tools are needed to analyze large-scale sequencing datasets. In particular, identification of copy number changes remains a challenging task due to their complexity, susceptibility to sequencing biases, variation in coverage data and dependence on genome-wide sample properties, such as tumor polyploidy, polyclonality in cancer samples, or frequency of de novo variation in germline genomes of pedigrees. The frequent need of core sequencing facilities to process samples from both normal and tumor sources favors multipurpose variant calling tools with functionality to process these diverse sets within a single software framework. This not only simplifies the overall bioinformatics workflow but also streamlines maintenance by shortening the software update cycle and requiring only limited staff training. Here we introduce Canvas, a tool for identification of copy number changes from diverse sequencing experiments including whole-genome matched tumor-normal, small pedigree, and single-sample normal resequencing, as well as whole-exome matched and unmatched tumor-normal studies. In addition to variant calling, Canvas infers genome-wide parameters such as cancer ploidy, purity, and heterogeneity. It provides fast and easy-to-run workflows that can scale to thousands of samples and can be easily incorporated into variant calling pipelines.

Comprehensive benchmarking of SNV callers for highly admixed tumor data.

Precision medicine attempts to individualize cancer therapy by matching tumor-specific genetic changes with effective targeted therapies. A crucial first step in this process is the reliable identification of cancer-relevant variants, which is considerably complicated by the impurity and heterogeneity of clinical tumor samples. We compared the impact of admixture of non-cancerous cells and low somatic allele frequencies on the sensitivity and precision of 19 state-of-the-art SNV callers. We studied both whole exome and targeted gene panel data and up to 13 distinct parameter configurations for each tool. We found vast differences among callers. Based on our comprehensive analyses we recommend joint tumor-normal calling with MuTect, EBCall or Strelka for whole exome somatic variant calling, and HaplotypeCaller or FreeBayes for whole exome germline calling. For targeted gene panel data on a single tumor sample, LoFreqStar performed best. We further found that tumor impurity and admixture had a negative impact on precision, and in particular, sensitivity in whole exome experiments. At admixture levels of 60% to 90% sometimes seen in pathological biopsies, sensitivity dropped significantly, even when variants were originally present in the tumor at 100% allele frequency. Sensitivity to low-frequency SNVs improved with targeted panel data, but whole exome data allowed more efficient identification of germline variants. Effective somatic variant calling requires high-quality pathological samples with minimal admixture, a consciously selected sequencing strategy, and the appropriate variant calling tool with settings optimized for the chosen type of data.

Improving somatic variant identification through integration of genome and exome data

Cost-effective high-throughput sequencing technologies, together with efficient mapping and variant calling tools, have made it possible to identify somatic variants for cancer study. However, integrating somatic variants from whole exome and whole genome studies poses a challenge to researchers as the variants identified by whole genome analysis may not be identified by whole exome analysis and vice versa. Simply taking the union or intersection of the results may lead to too many false positives or too many false negatives. To tackle this problem, we use machine learning models to integrate whole exome and whole genome calling results from two representative tools, VCMM (with the highest sensitivity but very low precision) and MuTect (with the highest precision). The evaluation results, based on both simulated and real data, show that our framework improves somatic variant calling, and is more accurate in identifying somatic variants than either individual method used alone or using variants identified from only whole genome data or only whole exome data. Using machine learning approach to combine results from multiple calling methods on multiple data platforms (e.g., genome and exome) enables more accurate identification of somatic variants.

Validation of a Next-Generation Sequencing Pipeline for the Molecular Diagnosis of Multiple Inherited Cancer Predisposing Syndromes

Despite the growing knowledge of the genetic background behind the cancers that occur in a context of hereditary predisposition, personal or family cancer history may not be clear enough to support directional gene testing. Defined targeted next-generation sequencing gene panels allow identification of the causative disease mutations of multigene syndromes and differential diagnosis for syndromes with phenotypically overlapping characteristics. Herein, we established a next-generation sequencing analysis pipeline for the molecular diagnosis of multiple inherited cancer predisposing syndromes using the commercially available target sequencing panel TruSight Cancer. To establish the analysis pipeline, we included 22 control samples with deleterious mutations covering all genes currently analyzed at our institution by standard Sanger sequencing. We tested the pipeline using 51 samples from patients with a clinical diagnosis of neurofibromatosis type 1 (NF1), 10 of which without previous molecular characterization of the causative NF1 mutations. We propose a thoroughly validated analysis pipeline that combines Isaac Enrichment, Burrows-Wheeler Aligner Enrichment, and NextGENe for the alignment and variant calling, and GeneticistAssistant for variant annotation and prioritization. This pipeline allowed the identification of disease-causing mutations in all 73 patients, including a large duplication of 37 bp in NF1. We show that high sensitivity and specificity can be achieved by using multiple bioinformatic tools for alignment and variant calling and careful variant filtering, having in mind the clinical question.

Sequence analysis of pooled bacterial samples enables identification of strain variation in group A streptococcus

Knowledge of the genomic variation among different strains of a pathogenic microbial species can help in selecting optimal candidates for diagnostic assays and vaccine development. Pooled sequencing (Pool-seq) is a cost effective approach for population level genetic studies that require large numbers of samples such as various strains of a microbe. To test the use of Pool-seq in identifying variation, we pooled DNA of 100 Streptococcus pyogenes strains of different emm types in two pools, each containing 50 strains. We used four variant calling tools (Freebayes, UnifiedGenotyper, SNVer, and SAMtools) and one emm1 strain, SF370, as a reference genome. In total 63719 SNPs and 164 INDELs were identified in the two pools concordantly by at least two of the tools. Majority of the variants (93.4%) from six individually sequenced strains used in the pools could be identified from the two pools and 72.3% and 97.4% of the variants in the pools could be mined from the analysis of the 44 complete Str. pyogenes genomes and 3407 sequence runs deposited in the European Nucleotide Archive respectively. We conclude that DNA sequencing of pooled samples of large numbers of bacterial strains is a robust, rapid and cost-efficient way to discover sequence variation.

Evaluating Variant Calling Tools for Non-Matched Next-Generation Sequencing Data

Valid variant calling results are crucial for the use of next-generation sequencing in clinical routine. However, there are numerous variant calling tools that usually differ in algorithms, filtering strategies, recommendations and thus, also in the output. We evaluated eight open-source tools regarding their ability to call single nucleotide variants and short indels with allelic frequencies as low as 1% in non-matched next-generation sequencing data: GATK HaplotypeCaller, Platypus, VarScan, LoFreq, FreeBayes, SNVer, SAMtools and VarDict. We analysed two real datasets from patients with myelodysplastic syndrome, covering 54 Illumina HiSeq samples and 111 Illumina NextSeq samples. Mutations were validated by re-sequencing on the same platform, on a different platform and expert based review. In addition we considered two simulated datasets with varying coverage and error profiles, covering 50 samples each. In all cases an identical target region consisting of 19 genes (42,322 bp) was analysed. Altogether, no tool succeeded in calling all mutations. High sensitivity was always accompanied by low precision. Influence of varying coverages- and background noise on variant calling was generally low. Taking everything into account, VarDict performed best. However, our results indicate that there is a need to improve reproducibility of the results in the context of multithreading.

Detailed simulation of cancer exome sequencing data reveals differences and common limitations of variant callers

BackgroundNext-generation sequencing of matched tumor and normal biopsy pairs has become a technology of paramount importance for precision cancer treatment. Sequencing costs have dropped tremendously, allowing the sequencing of the whole exome of tumors for just a fraction of the total treatment costs. However, clinicians and scientists cannot take full advantage of the generated data because the accuracy of analysis pipelines is limited. This particularly concerns the reliable identification of subclonal mutations in a cancer tissue sample with very low frequencies, which may be clinically relevant.ResultsUsing simulations based on kidney tumor data, we compared the performance of nine state-of-the-art variant callers, namely deepSNV, GATK HaplotypeCaller, GATK UnifiedGenotyper, JointSNVMix2, MuTect, SAMtools, SiNVICT, SomaticSniper, and VarScan2. The comparison was done as a function of variant allele frequencies and coverage. Our analysis revealed that deepSNV and JointSNVMix2 perform very well, especially in the low-frequency range. We attributed false positive and false negative calls of the nine tools to specific error sources and assigned them to processing steps of the pipeline. All of these errors can be expected to occur in real data sets. We found that modifying certain steps of the pipeline or parameters of the tools can lead to substantial improvements in performance. Furthermore, a novel integration strategy that combines the ranks of the variants yielded the best performance. More precisely, the rank-combination of deepSNV, JointSNVMix2, MuTect, SiNVICT and VarScan2 reached a sensitivity of 78% when fixing the precision at 90%, and outperformed all individual tools, where the maximum sensitivity was 71% with the same precision.ConclusionsThe choice of well-performing tools for alignment and variant calling is crucial for the correct interpretation of exome sequencing data obtained from mixed samples, and common pipelines are suboptimal. We were able to relate observed substantial differences in performance to the underlying statistical models of the tools, and to pinpoint the error sources of false positive and false negative calls. These findings might inspire new software developments that improve exome sequencing pipelines and further the field of precision cancer treatment.

STR-realigner: a realignment method for short tandem repeat regions.

BackgroundIn the estimation of repeat numbers in a short tandem repeat (STR) region from high-throughput sequencing data, two types of strategies are mainly taken: a strategy based on counting repeat patterns included in sequence reads spanning the region and a strategy based on estimating the difference between the actual insert size and the insert size inferred from paired-end reads. The quality of sequence alignment is crucial, especially in the former approaches although usual alignment methods have difficulty in STR regions due to insertions and deletions caused by the variations of repeat numbers.ResultsWe proposed a new dynamic programming based realignment method named STR-realigner that considers repeat patterns in STR regions as prior knowledge. By allowing the size change of repeat patterns with low penalty in STR regions, accurate realignment is expected. For the performance evaluation, publicly available STR variant calling tools were applied to three types of aligned reads: synthetically generated sequencing reads aligned with BWA-MEM, those realigned with STR-realigner, those realigned with ReviSTER, and those realigned with GATK IndelRealigner. From the comparison of root mean squared errors between estimated and true STR region size, the results for the dataset realigned with STR-realigner are better than those for other cases. For real data analysis, we used a real sequencing dataset from Illumina HiSeq 2000 for a parent-offspring trio. RepeatSeq and lobSTR were applied to the sequence reads for these individuals aligned with BWA-MEM, those realigned with STR-realigner, ReviSTER, and GATK IndelRealigner. STR-realigner shows the best performance in terms of consistency of the size of estimated STR regions in Mendelian inheritance. Root mean squared error values were also calculated from the comparison of these estimated results with STR region sizes obtained from high coverage PacBio sequencing data, and the results from the realigned sequencing data with STR-realigner showed the least (the best) root mean squared error value.ConclusionsThe effectiveness of the proposed realignment method for STR regions was verified from the comparison with an existing method on both simulation datasets and real whole genome sequencing dataset.

THapMix: simulating tumour samples through haplotype mixtures.

Large-scale rearrangements and copy number changes combined with different modes of clonal evolution create extensive somatic genome diversity, making it difficult to develop versatile and scalable variant calling tools and create well-calibrated benchmarks. We developed a new simulation framework tHapMix that enables the creation of tumour samples with different ploidy, purity and polyclonality features. It easily scales to simulation of hundreds of somatic genomes, while re-use of real read data preserves noise and biases present in sequencing platforms. We further demonstrate tHapMix utility by creating a simulated set of 140 somatic genomes and showing how it can be used in training and testing of somatic copy number variant calling tools. tHapMix is distributed under an open source license and can be downloaded from https://github.com/Illumina/tHapMix CONTACT: sivakhno@illumina.comSupplementary information: Supplementary data are available at Bioinformatics online.