Abstract 533: Accurately identifying expressed somatic variants for neoantigen detection and immuno-oncology

Accurate detection of somatic variants is a staple of both research and clinical cancer analysis, with applications ranging from detecting new common driver mutations in large patient cohorts to selecting therapeutic small molecule treatment courses for an individual patient. Recent research into neoantigens and immunotherapy has shown great promise as a precision therapeutic, and somatic variant detection by next-generation sequencing represents an ideal method of identifying candidate neoantigens. Somatic variant detection typically involves assaying the DNA for changes in gene sequences without assessing whether those variants are actually expressed in RNA. However, the expression of small variants is key because only expressed peptides will be displayed as neoantigens on the cell surface.

From a technical standpoint, detection of somatic variants in the RNA represents additional challenges above and beyond those of somatic detection in DNA. The widely varying expression levels of cancer genes, alternative splicing, and RNA editing are all features that make somatic variant calling in RNA uniquely challenging. However, accurately detecting variants directly from expressed transcripts is beneficial to neoantigen prediction, and therefore we sought to create and validate a method for somatic variant calling in RNA.

We have designed a highly accurate expression-based somatic variant detection pipeline utilizing extensive discovery and filtering methods to overcome the challenges inherent in RNA somatic variant calling. We validated our pipeline using a combination of well-characterized cell lines, commercially available reference standards, and real world FFPE patient samples. To our knowledge, this is the most extensive validation of its kind to date, representing over 29,158 small variants across 39 samples. In testing, we measured our detection method at >99% sensitivity and >99% PPV using a combination of gold set small variants and orthogonal validation. This method, in combination with our validated DNA somatic variant calling pipeline (>99% sensitivity and >99% PPV), enables precise detection of variant expression levels in a given sample, even at low allele frequency (5%).

After validating our RNA somatic variant calling method, we applied it to detect candidate neoantigens in patient tumor samples. We performed HLA typing for each sample using HLAssign software and predicted MHC presentation of the expressed somatic variants. In ongoing studies, we are validating our most promising putative neoantigens using orthogonal technologies and demonstrating our ability to detect the most promising clinically effective peptides for therapy.

Citation Format: Sean M. Boyle, Michael J. Clark, Ravi Alla, Shujun Luo, Deanna M. Church, Elena Helman, Parin Sripakdeevong, John West, Rich Chen. Accurately identifying expressed somatic variants for neoantigen detection and immuno-oncology. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 533.

Abstract 3169: The benefits and burdens of assaying matched normal tissue when sequencing cancer genomes

Targeted sequencing assays are increasingly used to identify tumor mutations that guide therapeutic decisions. Interpretation of a cancer variant's origin and therapeutic impact poses analytical challenges. Recent studies have indicated that jointly analyzing a tumor with its matched normal can accurately discriminate between tumor-specific (somatic) and inherited (germline) mutations. Moreover, a NHGRI/NCI Clinical Sequencing Exploratory Research Consortium Tumor Working Group just released a set of guidelines recommending that laboratories performing cancer sequencing tests should include germline variants. However, procurement of a matched sample is often logistically impractical. In the absence of a matched normal, large databases and analytical techniques are currently used to identify cancer variants in tumor sequencing data. Whether the benefits outweigh the additional burden of sequencing the matched normal for accurate detection of cancer-relevant mutations remains an open question.

To compare tumor-only and tumor/normal analysis of cancer samples, we collected a set of >100 formalin-fixed (FFPE) and fresh frozen cancer samples of various tumor types, where matched normal blood or adjacent tissue was available. We performed augmented target enrichment sequencing (exome and large cancer gene panel) of both DNA and RNA. The data was analyzed using cancer bioinformatics pipelines that detect base substitutions, small insertions/deletions, copy number alterations, and gene fusions in both tumor-only and tumor/normal modes. Variants were annotated using described clinical actionability filtering strategies. Analysis of germline variants for secondary findings was performed.

We find that 67% of mutations detected in tumor-only mode are reclassified as germline variants when analyzed together with the matched normal sample. These include mutations in hereditary cancer predisposition genes, such as BRCA1, VHL, and other genes with ACMG guidelines that warrant germline variant classification and appropriate management. Clinically actionable mutations may be miscalled as somatic when a matched normal is not available; however, we find the definition of ‘actionable’ can greatly impact the results of this analysis. Finally, the use of newly available large datasets, such as ExAC, substantially decreases the number of miscalled somatic variants in the absence of a matched normal.

The effects of administering targeted therapies to patients with germline mutations in the relevant gene are largely unknown. Mutations of putative germline origin may be important for hereditary cancer knowledge and tumor treatment, and should be reported as such. For NGS-based cancer interpretation to guide clinical decisions in a practical and cost-effective manner, highly optimized tumor-only and tumor/normal analyses must be available with proper attention to germline consent, classification and education.

Citation Format: Elena Helman, Michael J. Clark, Ravi Alla, Sean M. Boyle, Shujun Luo, Selene Virk, Deanna Church, Parin Sripakdeevong, Jason Harris, Mirian karbelashvili, Christian Haudenschild, John West, Richard Chen. The benefits and burdens of assaying matched normal tissue when sequencing cancer genomes. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3169.

Abstract 4744: Solving genomic assay trade-offs with an optimized, extended cancer gene panel for research and clinical applications

Choosing a genomic assay for cancer research is complicated by trade-offs. Cancer gene panels are a common choice, but they target mutational hotspots in a relatively small number of genes, often for cancers that are most commonly tested and that have common genetic etiologies. A common alternative is exome sequencing, which includes all the coding genes but, due to its larger genomic footprint, cannot reach the same depths as panels and therefore is less able to deal with low tumor purity and heterogeneity. Whole genome sequencing trades off very shallow depth and coverage over vast regions of uninterpretable genomic sequence in exchange for the identification of intergenic variants and structural variant breakpoints. All of these assays can be supplemented with RNA sequencing in order to capture gene fusions, allelic expression, splice isoforms, and gene expression. RNAseq comes with its own costs: the need to extract RNA from the same tissue, the need to perform a second assay, and the need to analyze a very different type of data from DNA sequencing.

The trade-offs generally come down to three major issues: depth of sequencing, specific genes targeted, and cost. To solve these, we designed an extended, optimized cancer gene panel facilitating high depth sequencing at low cost. We started by identifying a comprehensive list of over 1,300 cancer genes. These genes were chosen through exhaustive cancer gene database and literature curation, and include genes from all major cancer pathways and from the Cancer Gene Census. We then took this gene list and applied an augmented targeting design strategy that we have previously used to create an augmented exome enrichment platform which fills in gaps that standard technical exomes miss.

To validate the panel and analysis, we identified test samples including well-described cancer cell lines, cell line mixtures with engineered cancer variations, and formalin-fixed neoplastic tissues. We then performed a series of tests with these samples to measure the panel's small variant sensitivity and specificity, gauge its limits of detection, validate the detection of gene fusions, and demonstrate its ability to identify copy number alterations and loss of heterozygosity. In engineered cell lines, we detected 100% of small variants down to 5% allele frequency. We also mixed the cancer cell lines in various ratios and found similarly high sensitivity as well as very high specificity for small variant detection. We further compared our structural variation calls in the DNA and our fusion calls in the RNA with known data and found that we had very high concordance with known variations.

These studies demonstrate that an extended, augmented cancer gene panel strategy solves many genomic assay trade-offs and leads to high accuracy and variant yield for cancer research applications.

Citation Format: Michael J. Clark, Sean M. Boyle, Elena Helman, Shujun Luo, Gabor Bartha, Massimo Morra, Anil Patwardhan, Christian Haudenschild, Mirian Karbelashvili, Parin Sripakdeevong, Jason Harris, Deanna Church, Stephen Chervitz, John West, Richard Chen. Solving genomic assay trade-offs with an optimized, extended cancer gene panel for research and clinical applications. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4744. doi:10.1158/1538-7445.AM2015-4744

Abstract 3899: Integrating RNA/DNA analysis with a comprehensive cancer panel to improve interpretations of stage four metastatic renal cell carcinoma

Next-generation sequencing is now commonly applied to measure genomic modifications, providing the opportunity for a more accurate view of cancer progression. That being said, several challenges including tumor purity, heterogeneity and the broad set of cancer associated cellular functions, increase the difficulty of interpreting cancer genomes and make analyzing them distinct from analysis of germline diseases. The ability to detect low allelic mutations in the broad set of cancer-associated genes is critically important. To address these issues, we developed an extended cancer panel, a targeted enrichment sequencing platform that includes over 1,300 cancer genes and 200 miRNAs that improves coverage in traditionally difficult to sequence NGS regions. In this study, we apply an integrated RNA/DNA analysis approach as well as our extended cancer panel to explore mutations in metastatic renal cell carcinoma (RCC) tissue samples before and after treatment.

Our goal with this study was to test the importance of integrating DNA and RNA analysis, which allows for detection of mutational events that cannot be observed by either method alone. With RNA analysis it is possible to detect gene fusion events, expression of small variants (SNVs/indels), and gene expression levels while DNA analysis allows for detection of loss of heterozygosity (LOH), copy number variants (CNVs), and small variants. By integrating both the RNA and DNA analysis, it is possible to identify unique variant classes, such as unexpressed variants, allele specific expression (ASE), effects of CNVs on gene expression levels, and how fusion gene integration sites affect CNVs. In our study, each variant class was interpreted pre- and post-treatment, allowing for optimal analysis of treatment effectiveness.

By applying both our RNA/DNA integration approach and our extended cancer panel to investigate metastatic RCC samples, we were able to classify many genomic alterations that drive cancer progression. For example, we found that anywhere from 5% to 30% of small variant mutations in driver genes called in the DNA reside in genes that are unexpressed in the RNA. Likewise, we noted multiple cases of ASE, where high profile variants observed in DNA were entirely absent in the RNA despite high expression levels. Strong correlations between CNVs observed in DNA and gene expression changes found in RNA were also detected. We also observed that about half of the RCC samples we tested were biallelic for VHL mutations, suggesting inherited predisposition combined with second hit somatic mutations, facilitating cancer progression.

This analysis demonstrates how combining an integrated RNA/DNA approach with a cancer focused augmented enrichment panel allows for detection of both low allelic representation variants and unique variant classes, both of which are critical for accurate interpretation of cancer samples.

Citation Format: Sean M. Boyle, Michael J. Clark, Elena Helman, Alexander S. Parker, Thai Ho, Shujun Luo, Scott Kirk, Parin Sripakdeevong, Mirian Karbelashvili, Deanna M. Church, Michael Snyder, John West, Rich Chen. Integrating RNA/DNA analysis with a comprehensive cancer panel to improve interpretations of stage four metastatic renal cell carcinoma. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 3899. doi:10.1158/1538-7445.AM2015-3899

Structure determination of noncanonical RNA motifs guided by ¹H NMR chemical shifts

Structured non-coding RNAs underline fundamental cellular processes, but determining their 3D structures remains challenging. We demonstrate herein that integrating NMR ¹H chemical shift data with Rosetta de novo modeling can consistently return high-resolution RNA structures. On a benchmark set of 23 noncanonical RNA motifs, including 11 blind targets, Chemical-Shift-ROSETTA for RNA (CS-ROSETTA-RNA) recovered the experimental structures with high accuracy (0.6 to 2.0 Å all-heavy-atom rmsd) in 18 cases.

Serverification of molecular modeling applications: the Rosetta Online Server that Includes Everyone (ROSIE)

The Rosetta molecular modeling software package provides experimentally tested and rapidly evolving tools for the 3D structure prediction and high-resolution design of proteins, nucleic acids, and a growing number of non-natural polymers. Despite its free availability to academic users and improving documentation, use of Rosetta has largely remained confined to developers and their immediate collaborators due to the code's difficulty of use, the requirement for large computational resources, and the unavailability of servers for most of the Rosetta applications. Here, we present a unified web framework for Rosetta applications called ROSIE (Rosetta Online Server that Includes Everyone). ROSIE provides (a) a common user interface for Rosetta protocols, (b) a stable application programming interface for developers to add additional protocols, (c) a flexible back-end to allow leveraging of computer cluster resources shared by RosettaCommons member institutions, and (d) centralized administration by the RosettaCommons to ensure continuous maintenance. This paper describes the ROSIE server infrastructure, a step-by-step 'serverification' protocol for use by Rosetta developers, and the deployment of the first nine ROSIE applications by six separate developer teams: Docking, RNA de novo, ERRASER, Antibody, Sequence Tolerance, Supercharge, Beta peptide design, NCBB design, and VIP redesign. As illustrated by the number and diversity of these applications, ROSIE offers a general and speedy paradigm for serverification of Rosetta applications that incurs negligible cost to developers and lowers barriers to Rosetta use for the broader biological community. ROSIE is available at http://rosie.rosettacommons.org.

Correcting pervasive errors in RNA crystallography through enumerative structure prediction

Three-dimensional RNA models fitted into crystallographic density maps exhibit pervasive conformational ambiguities, geometric errors, and steric clashes. To address these problems, we present Enumerative Real-space Refinement ASsisted by Electron density under Rosetta (ERRASER), coupled to PHENIX (Python-based Hierarchical Environment for Integrated Xtallography) diffraction-based refinement. On 24 datasets, ERRASER automatically corrects the majority of MolProbity-assessed errors, improves average R_free factor, resolves functionally important discrepancies in non-canonical structure, and refines low-resolution models to better match higher resolution models.

RNA-Puzzles: a CASP-like evaluation of RNA three-dimensional structure prediction

We report the results of a first, collective, blind experiment in RNA three-dimensional (3D) structure prediction, encompassing three prediction puzzles. The goals are to assess the leading edge of RNA structure prediction techniques; compare existing methods and tools; and evaluate their relative strengths, weaknesses, and limitations in terms of sequence length and structural complexity. The results should give potential users insight into the suitability of available methods for different applications and facilitate efforts in the RNA structure prediction community in ongoing efforts to improve prediction tools. We also report the creation of an automated evaluation pipeline to facilitate the analysis of future RNA structure prediction exercises.