Somatic Variant Calling Research Articles

GeDi: applying suffix arrays to increase the repertoire of detectable SNVs in tumour genomes

BackgroundCurrent popular variant calling pipelines rely on the mapping coordinates of each input read to a reference genome in order to detect variants. Since reads deriving from variant loci that diverge in sequence substantially from the reference are often assigned incorrect mapping coordinates, variant calling pipelines that rely on mapping coordinates can exhibit reduced sensitivity.ResultsIn this work we present GeDi, a suffix array-based somatic single nucleotide variant (SNV) calling algorithm that does not rely on read mapping coordinates to detect SNVs and is therefore capable of reference-free and mapping-free SNV detection. GeDi executes with practical runtime and memory resource requirements, is capable of SNV detection at very low allele frequency (<1%), and detects SNVs with high sensitivity at complex variant loci, dramatically outperforming MuTect, a well-established pipeline.ConclusionBy designing novel suffix-array based SNV calling methods, we have developed a practical SNV calling software, GeDi, that can characterise SNVs at complex variant loci and at low allele frequency thus increasing the repertoire of detectable SNVs in tumour genomes. We expect GeDi to find use cases in targeted-deep sequencing analysis, and to serve as a replacement and improvement over previous suffix-array based SNV calling methods.

AMLVaran: a software approach to implement variant analysis of targeted NGS sequencing data in an oncological care setting

BackgroundNext-Generation Sequencing (NGS) enables large-scale and cost-effective sequencing of genetic samples in order to detect genetic variants. After successful use in research-oriented projects, NGS is now entering clinical practice. Consequently, variant analysis is increasingly important to facilitate a better understanding of disease entities and prognoses. Furthermore, variant calling allows to adapt and optimize specific treatments of individual patients, and thus is an integral part of personalized medicine.However, the analysis of NGS data typically requires a number of complex bioinformatics processing steps. A flexible and reliable software that combines the variant analysis process with a simple, user-friendly interface is therefore highly desirable, but still lacking.ResultsWith AMLVaran (AML Variant Analyzer), we present a web-based software, that covers the complete variant analysis workflow of targeted NGS samples. The software provides a generic pipeline that allows free choice of variant calling tools and a flexible language (SSDL) for filtering variant lists. AMLVaran’s interactive website presents comprehensive annotation data and includes curated information on relevant hotspot regions and driver mutations. A concise clinical report with rule-based diagnostic recommendations is generated.An AMLVaran configuration with eight variant calling tools and a complex scoring scheme, based on the somatic variant calling pipeline appreci8, was used to analyze three datasets from AML and MDS studies with 402 samples in total. Maximum sensitivity and positive predictive values were 1.0 and 0.96, respectively. The tool’s usability was found to be satisfactory by medical professionals.ConclusionCoverage analysis, reproducible variant filtering and software usability are important for clinical assessment of variants. AMLVaran performs reliable NGS variant analyses and generates reports fulfilling the requirements of a clinical setting. Due to its generic design, the software can easily be adapted for use with different targeted panels for other tumor entities, or even for whole-exome data. AMLVaran has been deployed to a public web server and is distributed with Docker scripts for local use.

An Individualized Approach for Somatic Variant Discovery.

Somatic variant callers identify mutations found within cancer genome sequencing data through mapping sequencing reads to a universal reference genome and inferring likelihoods from statistical models. False positives, however, are common among various tools as mismatches with the universal reference can also occur due to germline variants. Previous applications of personalized reference construction are not amenable with cancer genome analysis. Here, we describe an individualized approach for somatic variant discovery through the step-by-step usage of Personalized Reference Editor for Somatic Mutation discovery in cancer genomics (PRESM), a personalized reference editor for somatic mutation discovery in cancer genomes.

Detection of Oncogenic and Clinically Actionable Mutations in Cancer Genomes Critically Depends on Variant Calling Tools

The analysis of cancer genomes provides fundamental information about its aetiology, the processes driving cell transformation or potential treatments. While researchers and clinicians are often only interested in the identification of oncogenic mutations, actionable variants or mutational signatures, the first crucial step in the analysis of any tumor genome is the identification of somatic variants in cancer cells (i.e. those that have been acquired during their evolution). For that purpose, a wide range of computational tools have been developed in recent years to detect somatic mutations in sequencing data from tumors. While there have been some efforts to benchmark somatic variant calling tools and strategies, the extent to which variant calling decisions impact the results of downstream analyses of tumor genomes remains unknown. Here we present a study to elucidate whether different variant callers (MuSE, MuTect2, SomaticSniper, VarScan2) and strategies to combine them (Consensus and Union) lead to different results in three important analyses of cancer genomics data: identification of cancer driver genes, quantification of mutational signatures and detection of clinically actionable variants. To this end, we tested how the results of these three analyses varied depending on the somatic mutation caller or strategy in five different projects from The Cancer Genome Atlas (TCGA). Our results show that variant calling decisions have a significant impact on these analyses, creating important differences that could even impact treatment decisions for some patients. Moreover, the Consensus calling strategy to combine the output of multiple variant calling tools, a very widely used strategy by the research community, can lead to the loss of some cancer driver genes and actionable mutations. On the other hand, the Union seems to be a legit strategy for some downstream analyses with a robust performance. Overall, our results point to important differences in critical analyses of tumor sequencing data depending on variant calling and highlight the limitations of widespread practices within the cancer genomics community.

Comparison of somatic variant detection algorithms using Ion Torrent targeted deep sequencing data

BackgroundThe application of next-generation sequencing in cancer has revealed the genomic landscape of many tumour types and is nowadays routinely used in research and clinical settings. Multiple algorithms have been developed to detect somatic variation from sequencing data using either paired tumour-blood or tumour-only samples. Most of these methods have been developed and evaluated for the identification of somatic variation using Illumina sequencing datasets of moderate coverage. However, a comprehensive evaluation of somatic variant detection algorithms on Ion Torrent targeted deep sequencing data has not been performed.MethodsWe have applied three somatic detection algorithms, Torrent Variant Caller, MuTect2 and VarScan2, on a large cohort of ovarian cancer patients comprising of 208 paired tumour-blood samples and 253 tumour-only samples sequenced deeply on Ion Torrent Proton platform across 330 amplicons. Subsequently, the concordance and performance of the three somatic variant callers were assessed.ResultsWe have observed low concordance across the algorithms with only 0.5% of SNV and 0.02% of INDEL calls in common across all three methods. The intersection of all methods showed better performance when assessed using correlation with known mutational signatures, overlap with COSMIC variation and by examining the variant characteristics. The Torrent Variant Caller also performed well with the advantage of not eliminating a high number of variants that could lead to high type II error.ConclusionsOur results suggest that caution should be taken when applying state-of-the-art somatic variant algorithms to Ion Torrent targeted deep sequencing data. Better quality control procedures and strategies that combine results from multiple methods should ensure that higher accuracy is achieved. This is essential to ensure that results from bioinformatics pipelines using Ion Torrent deep sequencing can be robustly applied in cancer research and in the clinic.

Deep whole-genome sequencing of 3 cancer cell lines on 2 sequencing platforms

To test the performance of a new sequencing platform, develop an updated somatic calling pipeline and establish a reference for future benchmarking experiments, we performed whole-genome sequencing of 3 common cancer cell lines (COLO-829, HCC-1143 and HCC-1187) along with their matched normal cell lines to great sequencing depths (up to 278x coverage) on both Illumina HiSeqX and NovaSeq sequencing instruments. Somatic calling was generally consistent between the two platforms despite minor differences at the read level. We designed and implemented a novel pipeline for the analysis of tumor-normal samples, using multiple variant callers. We show that coupled with a high-confidence filtering strategy, the use of combination of tools improves the accuracy of somatic variant calling. We also demonstrate the utility of the dataset by creating an artificial purity ladder to evaluate the somatic pipeline and benchmark methods for estimating purity and ploidy from tumor-normal pairs. The data and results of the pipeline are made accessible to the cancer genomics community.

Prevalence of Inherited Cancer Predisposition Mutations in a Cohort of Older AML Patients Enrolled on the Beat AML Master Trial

Prevalence of Inherited Cancer Predisposition Mutations in a Cohort of Older AML Patients Enrolled on the Beat AML Master Trial

1438P - SomaticNET: Neural network evaluation of somatic mutations in cancer

BackgroundDNA sequencing to identify variants is becoming increasingly valuable in clinical settings; including matching patients to approved targeted therapies, immunotherapies, and/or clinical trials. However, accurate calling of genetic variants from sequencing still remains challenging. With little corroboration between the different tools available, patients are at risk of being treated with therapies that are unsuitable for their cancer. MethodsHere we present a novel machine learning based method for the accurate identification of somatic variants in cancer patient tumour samples, with a neural network architecture from encoded raw sequencing read information of tumour/normal sample pairings into an image, enabling it to classify whether a variant is germline, somatic, or sequencing error. The model was trained and tested on in-silico spike-in data using bam-surgeon, and then validated on a multi-cancer and multi-center dataset and benchmarked against industry standard variant callers. ResultsThe approach, called somaticNET, outperforms existing industry standard tools in sensitivity and specificity, achieving an AUROC of ∼1.00 on the bam-surgeon dataset and an AUROC of ∼0.99 on the multi-cancer multicenter dataset. The model also works faster than other variant callers, in minutes compared to hours. ConclusionsUsing the power of machine learning for accurate somatic variant calling can improve patient matching to approved therapies and clinical trials, thus ensuring patients are given the right therapy at the right time to treat their cancer. Legal entity responsible for the studyThe authors. FundingCambridge Cancer Genomics. DisclosureG. Dubourg-Felonneau: Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics. D. Rebergen: Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics. C. Parsons: Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics. H. Thompson: Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics. J.W. Cassidy: Leadership role, Shareholder / Stockholder / Stock options, Full / Part-time employment, Officer / Board of Directors: Cambridge Cancer Genomics. N. Patel: Leadership role, Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics. H.W. Clifford: Leadership role, Shareholder / Stockholder / Stock options, Full / Part-time employment: Cambridge Cancer Genomics.

Abstract 5108: Somatic small variant and copy number alteration calling with the Genome Analysis Toolkit

Abstract Somatic small mutations, SNVs or Indels, and copy number alterations are the two categories of mutations with the largest impact on cancer tumors. The Broad Institute has released somatic variant calling workflows for small mutations (M2) and copy number alterations (ModelSegments) based on the Genome Analysis Toolkit (GATK). The suite of workflows can call variants in capture or whole-genome sequencing data and will include functional annotations (Funcotator), such as protein change (for small variants) and impacted gene (for all variants). Common artifacts in sequencing data, such as those arising from oxidative DNA damage, FFPE/deamination, or mapping errors, are corrected automatically. Evaluation of the workflows is standardized and repeatable, which allows tracking of performance across versions, both detection performance (e.g. sensitivity, precision), as well as runtime performance (e.g. CPU and RAM usage). A matched normal is not required for a given tumor sample, since the workflows can leverage pre-processed panels of normals (PoNs). The workflows are freely available, are portable (i.e. can be run on local, on-prem, or cloud compute), are optimized for cost reduction, and can be tuned to optimally leverage available compute.The measured sensitivity of M2 was at least 0.93 for small somatic nucleotide variants (SNVs) and 0.83 for small insertions/deletions (Indels) on DREAM1, DREAM2, and DREAM3 challenges, and on a titrated mixture of germline samples (&amp;gt;=100x depth, AF = 0.2). The measured precision of M2 ranged from 0.91 to 0.98 on DREAM1, DREAM2, and DREAM3 for both SNVs and Indels. The false positive rate (FPR) of M2 was between 0.03 and 0.21 FP/Mb for SNVs, and between 0.0 and 0.1 FP/Mb for indels, on twelve paired, replicate normal-normal samples. The cost of the M2 workflow is about USD$1.15 for a pair of 35x WGS matched tumor-normal samples, using Google Cloud Compute, and required about 32 hours of CPU time on a single core with 3GB RAM. The measured sensitivity of ModelSegments was at least 0.91 for deletions and amplifications across three cohorts of TCGA whole-exome samples (Stomach adenocarcinoma N=39, Thyroid carcinoma N=50, and Lung adenocarcinoma N=60). The measured specificity for the same set of cohorts was at least 0.96 for both deletions and amplifications. All results reported here were using the corresponding SNP Array results as a truth set. GATK MS cost was approximately USD$0.65 on a 30x WGS pair using Google Cloud Compute and required about 6 hours of CPU time with a single core. The RAM usage was varied automatically in the workflow to minimize cost, but was in the range of 2-13GB. Citation Format: Lee Lichtenstein, Jonn Smith, David Benjamin, Aaron Chevalier, Kristian Cibulskis, Samuel K. Lee, Eric Banks. Somatic small variant and copy number alteration calling with the Genome Analysis Toolkit [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5108.

Abstract 2474: Automated somatic variant classifier to reduce false positives identified by tumor normal variant callers

Abstract Accurate identification of somatic mutations is key to targeted cancer therapies. However, due to the complexity of next-generation sequencing, there are many variables that give rise to sequencing and alignment artifacts. Most sequencing workflows using tumor normal paired samples rely on the intersection or union of mutations by more than one somatic variant caller such as Mutect2, Strelka, VarScan2, Muse, and SomaticSniper. Yet, they still contain a number of false positive and false negative calls. Therefore, there is a need for a manual review of variants using Integrative Genomics Viewer (IGV) to incorporate information that is difficult to take into account in automated workflows. To alleviate the burden of manual inspection, we present an automated classification method to pre classify detected somatic variants into true or false classes for further manual review. Methods: We created a false positive classifier that utilizes automated IGV visualization of somatic variants. Our somatic workflow involves tumor normal DNA sequencing, BWAMEM alignment, deduplication of reads via Picard, somatic calling using two somatic callers, followed by the union of the detected variants, SnpEff annotation, and extracting nonsilent variants with variant allele fractions greater than 5 percent. We used the data for 1,413 nonsilent variants from a set of 10 tumor normal pairs of various cancer types. The variants were scored by cancer analysts using IGV. Initial false positive rate was 28 percent. We fine tuned Residual Deep Neural Network which utilizes weights from the network with the same architecture trained on ImageNet. Discriminative fine tuning was performed in two stages: first unfreezing the last fully connected layer only, and then unfreezing all the layers and training with differential learning rates. Results and Conclusion: Our method achieves 93 percent overall accuracy. The classifier greatly reduces the false positive rate for the set of variant calls from 28 percent down to 4 percent at the expense of calling 3 percent of true variants false. Previously, deep neural networks have been shown to yield the state of the art performance in multiple domains including image processing, speech, and text processing. In this work, we describe a way to fine tune Residual Neural Network trained on a large image data set to perform false positive variant detection in tumor-normal variant alignment snapshots from IGV. Our general approach that significantly reduces false positive rate of putative variants can be extended to any subset of somatic callers. Citation Format: Alena S. Harley, Corine K. Lau, Eve Shinbrot. Automated somatic variant classifier to reduce false positives identified by tumor normal variant callers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2474.

Abstract 1371: Assessing the utility of cell-free DNA in identifying prostate cancer and characterizing tumor heterogeneity via whole exome and whole genome, multi-region sequencing

Abstract Early cancer diagnosis, especially while the disease is still localized and before symptoms appear, results in significantly higher survival rates compared to late-stage diagnosis. At the time of diagnosis, it is also common to find multiple foci within a single prostate gland in men with localized disease. Cell-free DNA (cfDNA) may not only reflect underlying disease biology, but also simultaneously allow for the identification of genetically distinct tumor subclones. The objectives of this study are to determine if 1) cfDNA levels are able to distinguish between healthy individuals from patients with localized or metastatic castration-resistant prostate cancer (mCRPC), and 2) if somatic mutations identified in tumor tissue are detectable in cfDNA and representative of the distribution observed in tumor tissue. This study includes samples from 130 individuals at UCSF: 21 healthy donors, 100 patients with localized prostate cancer who underwent radical prostatectomy (RP), and 9 mCRPC patients. Blood samples and matched tissue from adjacent normal seminal vesicles and multiple tumor regions (1-9 samples per patient) were collected from patients undergoing radical prostatectomy. CfDNA was extracted from plasma, and the concentration and fragment length distribution were measured with a Bioanalyzer 2100. Comparisons of cfDNA levels between groups were assessed with a Welch’s t-test due to the potential for unequal variances. Fifty-seven samples from nine patients have been subjected to whole exome sequencing, and 22 samples from five patients have been subjected to whole genome sequencing at ~40x coverage. Somatic variant calling was performed with Broad Institute’s Firecloud platform (GATK4/MuTect2) for tumor tissue and with the Curio platform for cfDNA to build consensus sequences leveraging unique molecular tags. CfDNA levels are able to distinguish between healthy and localized (p = 0.005), as well as healthy and metastatic groups (p = 0.043). Tumor foci within a patient’s prostate gland are genetically heterogeneous, with the majority of somatic mutations private to tumor regions and a subset of mutations at the intersection of these regions. Preliminary analyses result in an average of 72 somatic SNVs and indels per tumor tissue region, with ~10% overlap between regions in the same patient. Additional mCRPC and follow-up blood samples are being collected from all patients. The association between cfDNA levels prior to RP surgery and biochemical recurrence will be investigated when follow-up collection concludes. Further analysis of mutational concordance, clonality, and copy number variation between tumor tissue DNA and cfDNA, along with clinical data, will be performed. Citation Format: Emmalyn Chen, Clinton Cario, Lancelote Leong, Karen Lopez, Patricia Li, Erica Oropeza, Imelda Tenggara, Janet Cowan, Jeffry Simko, Daniel Wells, Robin Kageyama, June Chan, Terence Friedlander, Pamela Paris, Peter Carroll, John Witte. Assessing the utility of cell-free DNA in identifying prostate cancer and characterizing tumor heterogeneity via whole exome and whole genome, multi-region sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1371.

Abstract 3666: The genomic landscape and clonal evolution of tumours arising in TP53 mutation carriers

Abstract Li-Fraumeni syndrome (LFS) is a familial cancer predisposition syndrome (CPS) caused by germline mutations in TP53, associated with a high frequency of sarcomas, breast cancers, adrenocortical carcinomas and CNS tumours. Recently, our group and others have identified distinct mutational signatures, defined by the spectrum and context of somatic mutations, in tumours arising in individuals with another CPS, constitutional mismatch repair deficiency (CMMRD). CMMRD tumours frequently harbor somatic mutations disrupting the proofreading function of the DNA polymerases POLE and POLD1, which consequently leads to an ultra-hypermutant cancer genome, with early MMR mutational signatures and late POLE/POLD1 mutational signatures. This ultrahypermutant tumor phenotype is essentially diagnostic for the syndrome and provides a rational for immune checkpoint inhibitors which have shown success in this context. Based on our results in CMMRD tumors we hypothesize that LFS tumours might likewise harbor distinct mutational events and/or evolutionary dynamics from sporadic tumours of the same histiotype. Although several cancer genomics landscape studies have included a handful of tumours from LFS patients, to our knowledge no study to date has attempted to comprehensively characterize the cancer genomes of LFS patients. To investigate the somatic mutational events driving tumourigenesis in LFS we performed whole-genome sequencing (WGS) analysis of 22 tumours derived from patients with pathogenic germline TP53 mutations. Tumours from germline TP53 wildtype patients were analyzed by the same methods to serve as a control data set. For each tumour sample, where possible, we performed WGS on multiple spatially distinct micro-dissected tumour regions in order to reconstruct the evolutionary history of each cancer. Somatic variant calling was performed at high sensitivity using in silico reconstructed high-depth bulk WGS (80-120X coverage), with somatic mutations, structural variants, copy number alterations, mutational signatures and subclones identified using MuTect2, delly, battenberg, SigProfiler and phyloWGS respectively. High confidence variants were identified using in-house designed filtering pipelines. The resulting analyses reveals the life history of LFS cancers is marked by a high frequency of early catastrophic genomic rearrangement events, a diverse range of somatic driver events and in at least some cases marked intratumoural spatial heterogeneity of CNVs and SNVs. Citation Format: Nicholas Light, Matthew Zatzman, Nathaniel Anderson, Vallijah Subasri, Mehdi Layeghifard, Ana Novokmet, James Tran, Richard de Borja, Fabio Fuligni, Joshua Schiffman, David Malkin, Adam Shlien. The genomic landscape and clonal evolution of tumours arising in TP53 mutation carriers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3666.

Abstract 3539: Impact of formalin time on targeted NGS performance in FFPE tissue

Abstract Targeted next generation sequencing (NGS) panel testing is now a standard of care for management of advanced stage cancer. At present, most targeted NGS oncology testing occurs with formalin fixed paraffin embedded (FFPE) tissue. This is due to the established utility and prevalence of FFPE in routine histologic classification, grading, staging, and ancillary testing. In addition, FFPE materials are an optimally suited medium for determining adequacy of the specimen prior to downstream NGS targeted panel testing. For these reasons, as well as a growing trend of minimally invasive procedures, clinical specimens are increasingly prioritized for FFPE to meet the bulk of standard oncology care. Currently, formalin fixation time guidelines do exist for specific immunohistochemical and in-situ hybridization oncology tests. However, specific FFPE processing guidelines have yet to be established that would optimize targeted NGS panel testing performance. In this study, our aim was to develop contrived FFPE materials using the FDA-led Sequencing Quality Control Phase II (SEQC2) consortium working group 2 (WG2) cell line materials. The overarching goal was to use these FFPE materials to identify an upper limit of formalin time associated with minimal impact on somatic variant calling performance in targeted NGS oncology panels. The SEQC2 - WG2 normal lymphoblast cell line was selected as a well-characterized test material to undergo systematic alteration in formalin fixation time (1, 2, 6 and 24 hours) prior to “routine” tissue processing and embedding (FFPE). These FFPE cell-block materials, derived from a single sample source (with both extracted FFPE DNA as well as shaved sections), were distributed to multiple labs running various targeted NGS amplicon and hybridization oncology panels. The expected outcome of these studies is to provide general guidance to the community on how formalin time in FFPE specimens impacts targeted NGS oncology panel testing performance. In addition, we hope to identify genomic regions (internal control regions) that serve as markers for the full range of formalin time FFPE-effect. This information will assist in improving the lower limit of detection for variant calling (&amp;lt;5% variant allele frequency). Citation Format: Thomas M. Blomquist, Erin Crawford, James Willey, Joshua Xu, Onco-panel Working Group of the Sequencing Quality Control Phase 2 (SEQC2) Consortium. Impact of formalin time on targeted NGS performance in FFPE tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3539.

Abstract 2220: Benchmarking of somatic variant calling algorithms for detection of circulating tumor DNA

Abstract Purpose: Establishing a reliable bioinformatics variant calling workflow is crucial for untargeted searches of low-frequency variants in plasma circulating cell-free DNA (ccfDNA). We sought to benchmark the performance of nine variant calling algorithms on simulated data and cancer patient ccfDNA sequencing samples. Methods: A simulated tumor dataset was constructed by in-silico introduction of pathogenic variants (303 SNVs and 91 INDELs; VAF of 0.1-12%) into a healthy control sequencing sample (128 gene capture panel; 128kb; 4200X depth). An unspiked replicate served as a control. Tumor and control datasets were analyzed by nine somatic variant calling algorithms available in the public domain (Table 1). Sensitivity and specificity were calculated for each caller. Variant calling algorithms that proved most successful in synthetic data (Lofreq, Mutect) were applied to 24 EGFR-mutant lung cancer patient ccfDNA library samples captured and sequenced under conditions identical to the benchmarking data set. Droplet digital PCR (ddPCR) was used to independently determine the VAFs of two EGFR SNVs (T790M and L858R) and exon 19 deletions in all lung cancer libraries (detected range: 0.1-48.5%). Results: Table 1:Sensitivity (Sens.) and false discovery rate (FDR) as determined for 9 variant callers. For synthetic data, results are shown for 1.6% VAF (NC = no call)CallerSynthetic data (1.6% VAF)Lung cancer ccfDNASNVINDELSNVINDEL% Sens.% FDR% Sens.% FDR% Sens. &amp;lt;1% VAF% Sens. &amp;gt;1% VAF% Sens. &amp;lt;1% VAF% Sens. &amp;gt;1% VAFLofreq87.15.792.35.675.094.466.7100Mutect47.52.291.26.762.588.916.787.5Mutect235.63.296.71.1Strelka17.83.57.69.2FreebayesNCNCNCNCVarscanNCNCNCNCVardictNCNCNCNCLumpy13.986.17.792.1DellyNCNCNCNC Conclusion: Performance of variant callers differed greatly, with Lofreq providing the best sensitivity at low VAF. In cancer ccfDNA libraries Lofreq detected most but not all low-frequency variants, indicating that ccfDNA diagnostics require optimized variant calling workflows to improve detection of variants with a VAF &amp;lt;1%. Citation Format: Preetida J. Bhetariya, Sabine Hellwig, David A. Nix, Gabor T. Marth, Mary P. Bronner, Hunter R. Underhill. Benchmarking of somatic variant calling algorithms for detection of circulating tumor DNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2220.

Abstract 2487: Application of RegTools to TCGA samples for the identification of tumor-specific splice variants

Abstract The interpretation of variants in cancer is frequently focused on direct protein coding alterations. However, this analysis strategy excludes somatic mutations in non-coding regions of the genome and even exonic mutations may have unidentified non-coding consequences. To address this issue, we created RegTools (Feng et al. 2018), a free, open-source software package that integrates analysis of somatic variant calls from genomic data with splice junctions extracted from transcriptomic data in order to efficiently identify variants that may cause aberrant splicing in tumors. Initially, this software was developed using 188 cancer samples from 4 cancer types and enabled us to identify important non-coding variants which would otherwise be undervalued or perhaps even completely ignored by traditional methods and annotators. Here, we present the application of RegTools to over 10,000 TCGA samples in an effort to identify somatic variants that are associated with alternative splicing patterns within these tumors. In order to perform this analysis, we obtained all tumor samples that had somatic variant VCF files and RNA-seq bam files from the Genomic Data Commons (GDC). Included within these samples are more than 700 solid tissue normal samples, which will be used as a panel of normals in order to filter out variants and splicing events that are not tumor specific. We plan to present on the significance of our findings when analyzing these samples within their respective tumor type and across all tumor types. We also plan to present on whether these aberrant splicing events that result from somatic mutations might interfere with predicted neoantigens pVAC-Seq (Hundal et al. 2016). RegTools is freely available and open source (www.regtools.org). Citation Format: Kelsy C. Cotto, Yang-Yang Feng, Zachary L. Skidmore, Obi L. Griffith, Malachi Griffith. Application of RegTools to TCGA samples for the identification of tumor-specific splice variants [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2487.

Abstract 1678: Accurate detection of expressed variation in RNA-seq

Abstract Somatic variant detection has long been an area of great interest and critical importance in cancer research and treatment. DNA-Seq based somatic variant calling pipelines have been maturing over the last several years and although there is still room for improvement, several methods have been developed capable of accurately detecting somatic variants from matched tumor/normal DNA. RNA-Seq has become a widely used tool for measuring gene and isoform abundance, identifying alternative splicing, and detection of gene fusions. Detection of expressed variation is receiving growing interest, however computational methods to detect this variation have received far less attention than that of DNA-Seq. We have developed an optimized RNA-Seq pipeline based upon the ABRA2 realigner capable of accurately detecting expressed somatic variation in RNA-Seq. We applied this pipeline to the TCGA Breast Cancer dataset. To assess the impact of identification of expressed variation as an indicator of variant significance, we compared the expression state of the top 20 cohort wide significantly mutated genes (SMG) as identified by The Cancer Genome Atlas Network (2012) with all other genes. At the median for SMGs, 23 somatic variants are expressed with over 90% of variants detected in DNA being expressed. For non-SMGs, the median number of expressed variants is 1 and the median fraction of expressed variants is 50%. Among non-SMGs, a total of six genes exceed the median total number of expressed variants and fraction of expressed variants for SMGs (AHNAK, CHD4, ERBB2, FASN, FOXA1 and MYH9). All of these variants have been previously implicated in breast cancer, however none of these variants were among the cohort wide SMGs identified in the original study and only FOXA1 was identified as being a subtype specific SMG (ER+,ER+/HER2-). Notably, variants in the massive TTN gene are found to be coding 73% of the time, but expressed in only 8% of cases. This is substantially less than the 90% found in SMGs and a likely indicator that TTN mutations are passenger variants. The median and third quartile RNA variant allele frequency (VAF) for SMGs is .41 and .69 versus .32 and .46 for DNA. By comparison, the median and third quartile VAF for non-SMGS was 0 and .27 in RNA and .20 and .30 for DNA. The increased VAF for RNA SMGs is a likely indicator of selection and can potentially further be used to identify variants of significance. We believe these results demonstrate the utility of variant expression as a potentail tool to aid in assessment of variant significance. Citation Format: Lisle E. Mose, David Marron, Joel S. Parker. Accurate detection of expressed variation in RNA-seq [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1678.

PS942 CD19 MUTATION FREQUENCY AFTER BLINATUMOMAB TREATMENT IS LOWER THAN AFTER CAR T‐CELLS TREATMENT

Background:A significant antileukemic activity of CD19‐directed T‐cell immunotherapies, namely chimeric antigen receptor (CAR) T‐cells and Blinatumomab (Blin), has been demonstrated in advanced B‐cell precursor acute lymphoblastic leukemia (BCP‐ALL). However, despite high response rate, up to 30% of patients develop relapse with loss of CD19 antigen on leukemic cells (Ruella, JCI, 2016). Due to the novelty of this escape mechanism, only few studies investigating its genetic cause have been published so far.Aims:Our aim was to identify somatic mutations acquired at the time of CD19‐negative relapse in patients after Blin treatment, employing exome sequencing of remission (rem), screening (scr) and relapse (rel) samples.Methods:Among patients with relapse/refractory or minimal residual disease (MRD)‐positive BCP‐ALL treated within MT103–203, MT103–206, MT103–211, and TOWER trials, 17 patients suffering a CD19‐negative relapse with sufficient DNA available were identified. Rem, scr and rel samples (median infiltration at scr 40%, range 6–95%; at rel 63%, 6–95%) from these patients were analysed with Nextera Exome Enrichment or IDT xGEN Exome Panel on an Illumina HiSeq 4000 sequencer with 2x 75 bp reads and 6 samples per sequencing lane, resulting in 80X‐90X on‐target average coverage. Two scr and five rel samples with IG/TR‐based MRD below 10% were sequenced with double coverage. Somatic variant calling was done by the bcbio‐nextgen software. Variants with abundance over 1/5 of the sample infiltration were marked as dominant (dom), otherwise marked as subclonal (sub).Results:Mutations of the CD19 gene affecting the final protein were identified in only 5/17 (29.4%) patients (Table 1). Two of the mutations were frameshift variants producing a premature termination codon, 2 were nonsense variants, and 1 of them was located in the splicing region. In 4/5 patients loss of heterozygosity was present in the CD19 region, leaving no functional allele. The remaining patient showed inconclusive results, likely due to a very low leukemic infiltration (9%).We further assessed all somatic variants present in the rel sample but not in the scr sample (i.e. when &lt;5 alternative reads). On average, 10 (range 1–40) dom mutations and 7 (range 0–29) sub mutations were identified per patient. The difference between samples with and without a CD19 mutation was not significant. No mutation was shared among different patients and only ABCE1 and KMT2D genes were recurrently mutated – each of them in 2 different patients without a CD19 mutation.Summary/Conclusion:Mutations with an effect on CD19 protein were identified in less than 1/3 of patients in our cohort. In contrast to our findings, a recent study identified truncating CD19 mutations in 12/12 patients with a CD19‐negative relapse after CAR T‐cell treatment (Orlando, Nature Medicine, 2018). This difference could be attributed to a stronger selection pressure of CAR T‐cell treatment compared to Blin, where also lower incidence of CD19‐negative relapses has been reported (Ruella, Comput Struct Biotechnol J., 2016).imageA genetic explanation for the loss of CD19 in the remaining patients (n = 12) of our cohort is currently lacking. Clustering of all mutations into pathways, as well as analysis of a control cohort of patients with relapse on standard of care chemotherapy are ongoing with the aim to identify other molecular pathways responsible for the CD19 loss. RNA is not available for this retrospective cohort of patients, therefore studying alternative splicing via transcriptome analysis is not possible.

PF362 A FIRST HIGH-DEFINITION LANDSCAPE OF SOMATIC MUTATIONS IN CHRONIC LYMPHOPROLIFERATIVE DISORDER OF NK CELLS

Background: Chronic Lymphoproliferative Disorder of NK cells (CLPD-NK) is included in the 2017 WHO classification of mature T- and NK-cell neoplasms as provisional entity. Similarly to T-cell large granular lymphocytes leukemia (T-LGLL), CLPD-NK is characterized by an indolent clinical course and is clearly distinguished from the aggressive NK cell leukemia (ANKL), this latter behaving as an acute severe disease. CLPD-NK and T-LGLL patients mostly present cytopenias when symptomatic. Accumulating evidence on the genomic landscape of T-LGLL pointed to a key role of somatic mutations within JAK-STAT (mostly in STAT3 and STAT5B genes) and RAS-MAPK pathways (Coppe et al. Leukemia 2017) in this form. Recent data highlighted more heterogeneous molecular features in ANKL, involving JAK2 and STAT3 mutations, lesions of epigenetic modifiers, tumour suppressors and a putative role of Epstein Barr Virus (Dufva et al. Nat Comm. 2018). Aims: In our cohort of 57 CLPD-NK patients the frequency of STAT mutations has been found significantly lower (<10% of all cases) compared to literature data (Barilà et al. Blood Cancer J. 2018). The genomic landscape of CLPD-NK patients negative for STAT mutations was thus investigated to better elucidate the molecular basis of the disease. Methods: Ten CLPD-NK patients were recruited according to negativity for STAT mutations and typical immunophenotype (CD16+/CD57±/CD56±). WES profiling (Agilent SureSelect 60 Mbp, Illumina sequencing, paired end reads) of immunomagnetically purified leukemic clone (>95% purity) and of normal granulocytes, as control, was obtained for each patient. Sequencing data were analyzed by an in-house bioinformatic pipeline encompassing data cleaning, read mapping against the reference genome and somatic variant calling with complementary approaches (MuTect, MuTect2 and Strelka2). Variants were next filtered according to population allele frequency (gnomAD <5%), clinical significance (COSMIC and ClinVar) and predicted functional impact on proteins (SnpEff, MetaSVM and MetaLR). Results: High sequencing depth was obtained with a mean coverage of 150 reads. From over 6,000 somatic variants detected in the cohort, 869 SNPs and 14 indels, rare and of high predicted impact were prioritized. Number of variants (88 per patient in average) and mutation spectra were very homogeneous in the cohort. Of note, the high discovery power of our settings disclosed an extensive subclonality of somatic variants, with over 78% with an allele variant frequency (VAF) lower than 5% in the leukemic clone (Figure 1A). Nevertheless, 20 deleterious and high VAF somatic variants occurred in each patient in average (Figure 1B). Very low recurrence of mutated genes in CLPD-NK patients was found. Pathway-derived mutation network analysis (Lovisa et al. Haematologica 2019) revealed functional connections among genes mutated in different patients. A major involvement of JAK-STAT pathway is likely to be ruled out, but all cases presented deleterious high VAF mutations of genes and pathways potentially impacting on cell survival, proliferation and chromatin-remodelling. In addition, somatic lesions in proteins involved in innate immunity and NK-cells activation seem particularly interesting.Summary/Conclusion: Ongoing comparison of genes and pathways mutated in CLPD-NK with available data for T-LGLL and ANKL is providing new clues on biological similarities and differences of these clinical entities. This study contributes to a better definition of the molecular mechanisms accounting for CLPD-NK, thus opening new areas of investigation.

NeoMutate: an ensemble machine learning framework for the prediction of somatic mutations in cancer

BackgroundThe accurate screening of tumor genomic landscapes for somatic mutations using high-throughput sequencing involves a crucial step in precise clinical diagnosis and targeted therapy. However, the complex inherent features of cancer tissue, especially, tumor genetic intra-heterogeneity coupled with the problem of sequencing and alignment artifacts, makes somatic variant calling a challenging task. Current variant filtering strategies, such as rule-based filtering and consensus voting of different algorithms, have previously helped to increase specificity, although comes at the cost of sensitivity.MethodsIn light of this, we have developed the NeoMutate framework which incorporates 7 supervised machine learning (ML) algorithms to exploit the strengths of multiple variant callers, using a non-redundant set of biological and sequence features. We benchmarked NeoMutate by simulating more than 10,000 bona fide cancer-related mutations into three well-characterized Genome in a Bottle (GIAB) reference samples.ResultsA robust and exhaustive evaluation of NeoMutate’s performance based on 5-fold cross validation experiments, in addition to 3 independent tests, demonstrated a substantially improved variant detection accuracy compared to any of its individual composite variant callers and consensus calling of multiple tools.ConclusionsWe show here that integrating multiple tools in an ensemble ML layer optimizes somatic variant detection rates, leading to a potentially improved variant selection framework for the diagnosis and treatment of cancer.

Conbase: a software for unsupervised discovery of clonal somatic mutations in single cells through read phasing

Accurate variant calling and genotyping represent major limiting factors for downstream applications of single-cell genomics. Here, we report Conbase for the identification of somatic mutations in single-cell DNA sequencing data. Conbase leverages phased read data from multiple samples in a dataset to achieve increased confidence in somatic variant calls and genotype predictions. Comparing the performance of Conbase to three other methods, we find that Conbase performs best in terms of false discovery rate and specificity and provides superior robustness on simulated data, in vitro expanded fibroblasts and clonal lymphocyte populations isolated directly from a healthy human donor.