Search

Changes in DNA copy number, i.e., somatic CNVs, are common genetic aberrations in cancers. The effects of CNV include alteration in gene dosage across large segments of the cancer genome affecting the expression of cancer driver genes by amplifications, or cancer suppressor genes by deletions. In addition, CNVs are markers of underlying rearrangements within or between chromosomes and there is increasing evidence supporting a greater role for CNVs in developing and maintaining neoplastic cell population diversity. Copy number aberrations can be estimated from next generation sequencing data, with high sensitivity and genomic resolution by sequencing the whole genome (WGS). For this study, we demonstrated that high quality CNV calls can be extracted in a fast and cost-effective way from low-coverage whole genome sequencing. Novaseq S2 flowcells (Illumina Inc) enables to obtain an average coverage of 3-4x per sample after pooling up to 96 samples per flowcell. We examined three different copy number detection tools (CNVkit, BicSeq, and seqCBS) from paired tumor and normal WGS using microarray data as a reference. Pearson correlations were computed between the reference and CNVs from the WGS in two fashion; i) segment based and ii) gene based. The segment based comparison used sliding window of 100 K bp while gene based comparison used segments at the gene level. We found high correlations between microarray and WGS segments. The highest correlations were obtained by CNVkit, ranging from 0.964 to 0.985 (SD: 0.973 - 0.007) and BicSeq, ranging from 0.963 to 0.986 (SD: 0.975 - 0.008). These results open the prospect of assessing large cancer cohorts of hundreds of samples at a reasonable cost. We are planning to apply this method to a large cohort of Stage III colon cancer patients and determine the clinical relevance of CNVs for survival. Citation Format: HoJoon Lee, Li Charlie Xia, Stephanie Greer, John Bell, Sue M. Grimes, Christina Wood Bouwens, Giwon Shin, Billy TC Lau, Lucas Johnson, Noemi Andor, Kenneth Day, Mickey Miller, Helaman Escobar, Lincoln Nadauld, Hanlee P. Ji, Paul Van Hummelen. High-quality CNV segments from low-coverage whole genome sequencing from FFPE cancer biopsies based on an evaluation of multiple CNV tools [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 438.

Formalin-fixed, paraffin-embedded (FFPE) storage is a universally-adopted, cost-effective and long-term solution for tissue storage. Given the abundance of FFPE clinical samples, they provide a good source of tissue for genome sequencing studies. However, the extraction of nucleic acids from FFPE tissues yield highly fragmented DNA of variable quality. The poor quality and low yield make it difficult to generate adequate sequencing libraries for genomic studies. For a population cancer genome study, we optimized a procedure for whole-genome sequencing on FFPE tumor and normal tissue of colorectal cancer patients. This pilot study aimed to first identify changes in DNA copy number using low-pass sequencing, with the additional goal of later adding structural variant and gene fusion detection and single nucleotide variant analysis with deeper sequencing. We processed, sequenced and analyzed 193 patients. These FFPE samples had varying storage characteristics. We optimized the initial extraction quality thresholds, DNA extraction and library preparation to generate high quality sequencing data. We identified an initial sample quality and library threshold to hold the best correlation for sequencing quality. Optimal yield and DNA size from extracted FFPE was achieved by using Promega's LEV DNA FFPE Extraction Kit. Library prep optimization involved using mechanical shearing of DNA to increase overall library size for structural variant detection as well as using unique dual index adapters to identify index swapping artifacts during sequencing. An additional cleanup was also performed post-fragmentation to increase overall library size. Overall, we successfully developed a robust protocol for conducting whole genome sequencing on FFPE tumor samples. Citation Format: Matthew Kubit, Christina Wood-Bouwens, Sue Grimes, John Bell, GiWon Shin, Billy Lau, Mickey Miller, Kenneth Day, Helaman Escobar, Hanlee Ji, Lincoln Nadauld, Paul Van Hummelen. High-throughput whole-genome sequencing of formalin fixed, paraffin-embedded tissues from colorectal cancer patients [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 4331.

Viruses are a major contributor to oncogenesis, causing 10-15% of human cancers. Molecular pathways involved in malignant transformation are frequently activated by genetic alterations, including but not limited to, somatic mutations, copy number aberrations, structural variants, and oncoviruses. Precision cancer medicine aims to classify tumors by site, histology, and molecular tests to determine an “individualized” profile of cancer alterations. However, clinical tests for these various alterations are sequential, time consuming, and use a lot of material, which is often quite limited (e.g., biopsies). Moreover, tests for the presence of viral sequence are generally performed separately to tests (such as massively parallel sequencing) to detect human genomic alterations. Here we present a hybrid capture and massively parallel sequencing approach to detect viral infection concurrently with targeted genomic analysis, which may decrease assay costs, increase sensitivity and scalability, and detect many types of alterations, thereby providing a more complete tumor genetic profile all from a single sample. We have created a custom hybrid capture probeset for targeted Illumina sequencing to determine whether oncoviruses are present in tissue samples and also determine if the virus has integrated into the host’s genome. We have created both ‘detection’ and ‘integration’ baits for several oncoviruses, including polyomaviruses, human papilloma viruses, Epstein-Barr virus, human cytomegalovirus, Kaposi sarcoma herpesvirus, human T-lymphotropic virus, and hepatitis B virus. To distinguish between different strains of a single virus, strain-specific detection baits were created to bind to variable regions of viral genomes. The integration bait was designed to bind to regions of the viral genomes that are commonly integrated into the human genome. This baitset can also be combined with other capture panels targeting oncogenes to simultaneously determine infection and integration statuses, as well as somatic mutations, copy number and structural variants. To detect virus presence, reads were aligned to a hybrid reference of both the human, and targeted virus genomes. Viral integration status and integration loci were determined by leveraging discordant read pairs that aligned to both the human genome and a viral genome. We have tested our techniques on tissue samples that were infected with either Merkel Cell Polyomavirus or Epstein-Barr virus, as determined using quantitative polymerase chain reaction (qPCR) or immunohistochemistry (IHC) techniques, and have successfully detected these viruses and identified viral integration loci. Overall, this viral hybrid capture probeset provides the ability to simultaneously determine a tissue sample’s infection and viral integration status alongside other somatic genomic analyses, saving both time and sample material. Citation Format: Robert T. Burns, Samuel S. Hunter, Matthew D. Ducar, Aaron R. Thorner, James A. Decaprio, Paul Van Hummelen, Alexander Frieden, Anwesha Nag, Haley A. Coleman, Michael K. Slevin, Andrea Clapp, Samantha D. Drinan, Suzanne R. McShane, Edwin Thai, Priyanka Shivdasani, Joshua Bohannon, Johann Hoeftberger, Reuben Jacobs, Bruce M. Wollison, Neil A. Patel, Monica D. Manam, Phani Davineni, Matthew Meyerson, Laura E. MacConaill. Oncovirus detection and integration analysis from human tumor samples using targeted massively parallel sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 562. doi:10.1158/1538-7445.AM2017-562

Massively parallel sequencing (MPS) is increasingly used in the laboratory and clinic to identify genomic factors contributing to tumorigenesis, such as somatic mutations, DNA insertions and deletions, transcriptome and epigenetic changes, and chromosomal abnormalities. Because tumor specimens are often quite limited or are from rare and precious samples, it is necessary to prepare sequencing libraries from minimal amounts of DNA. Indeed, it can be challenging to extract sufficient amounts of starting material from fresh-frozen paraffin embedded (FFPE) samples for library construction, and frequently samples are not processed because they do not reach minimum requirements. Therefore, it is of great importance to improve currently available protocols to process low yield FFPE-derived DNA for MPS. This will allow for the inclusion of more tumor samples to be sequenced per cohort, resulting in increased ability to identify genomic factors involved in tumorigenesis. Here, we optimized the KAPA Library Preparation Kit (Kapa Biosystems, Wilmington, MA) to successfully perform library construction from low-yield FFPE and fresh frozen samples. The kit normally requires an input of at least 100 ng of DNA. However, we were able to modify the protocol to prepare libraries from as little as 10 ng of starting material by optimizing the ratio of adapter:input DNA and SPRI-clean-up, performing low-volume reactions, and using IDT universal blockers during hybrid capture. Additional library enrichment PCR was not required. Library yields were sufficient for downstream hybrid capture and sequencing, and sequencing metrics were comparable to samples that were prepped using the manufacturer's recommendations. Experiments are underway to demonstrate that library complexity remains unchanged. Citation Format: Aaron R. Thorner, Liuda Ziaugra, Matthew D. Ducar, Ling Lin, Angelica Laing, Haley A. Coleman, Suzanne R. McShane, Andrea Clapp, Rachel R. Paquette, Bruce M. Wollison, Johann Hoeftberger, Neil A. Patel, Samuel S. Hunter, Monica D. Manam, Laura E. MacConaill, William C. Hahn, Matthew L. Meyerson, Paul van Hummelen. Optimization of library construction for massively parallel sequencing using low-input, FFPE-derived DNA without additional PCR amplification. [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 3644.

Of pediatric solid tumors, as many as 10% of tumors are categorized as rare. Many of these rare tumors lack standard effective known therapy. The ability to identify vulnerabilities for many rare tumors has been significantly limited by the lack of in vitro and in vivo models. Furthermore, current approaches to study such vulnerabilities are usually limited to a specific compound or target. Our objectives were 1) to develop a platform to collect tumor samples and generate in vitro models and 2) to develop systematic and orthogonal approaches focused on currently known druggable cancer targets to identify vulnerabilities in these difficult to treat cancers. We have developed a proof of concept cell line from a patient who succumbed to progressive undifferentiated sarcoma treated on an aggressive multi-therapy regimen. This cell line, in its early passages, has novel gene fusions that match that of the primary tumor. Furthermore, even at early passages, this cell line was amenable to high throughput functional screens. Using a targeted pooled shRNA screen (employing matched seed controls) and an analogous CRISPR screen we identified dependencies to XPO1 and CDK4. In parallel, compounds against these targets were identified in a small molecule compound screen. These targetable dependencies were further validated in vivo with a micro-dosing device. These observations identify new targets in this rare malignancy. Furthermore, this suggests that the interrogation of patient derived cell lines facilitates the identification of testable therapeutic approaches. Citation Format: Andrew L. Hong, Glenn S. Cowley, Yuen-Yi Tseng, Jaime H. Cheah, Oliver Jonas, Mihir B. Doshi, Bryan D. Kynnap, Coyin Oh, Stephanie Meyer, Paul Clemons, Michael Burger, Francisca Vazquez, Barbara Weir, Gregory V. Kryukov, Alanna Church, Alma Imamovic, Aviad Tsherniak, Craig Bielski, Brian Crompton, Elizabeth Mullen, Charles Roberts, Carlos Rodriguez-Galindo, Katherine A. Janeway, Kimberly Stegmaier, Paul van Hummelen, Robert Langer, Levi A. Garraway, Stuart L. Schreiber, David E. Root, Jesse S. Boehm, William C. Hahn. Developing a functional genomics platform to interrogate rare pediatric cancers. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Pediatric Cancer Research: From Mechanisms and Models to Treatment and Survivorship; 2015 Nov 9-12; Fort Lauderdale, FL. Philadelphia (PA): AACR; Cancer Res 2016;76(5 Suppl):Abstract nr B38.

Background: Brain metastases are associated with a dismal prognosis. There is a limited understanding of the oncogenic alterations harbored by brain metastases and whether these are shared with their primary tumors. Our objectives were to (1) elucidate the evolutionary patterns leading to brain metastases and (2) identify whether brain metastases are genetically distinct from their primary tumors and other distal metastatic sites. Methods: We performed whole-exome sequencing of 104 matched brain metastases, primary tumors and normal tissue, including 7 cases with spatially and temporally separated brain metastasis sites and 8 cases with additional extracranial disease sites, including regional lymph nodes, and distal metastases. We developed novel computational tools to perform an integrative analysis of somatic mutations and copy-number alterations. This analysis allowed us to estimate the clonal architecture of the primary and metastases, and to reconstruct a phylogenetic tree relating the subclones from each patient. Results: In all related cancer samples, we observed branched evolution, where all metastatic and primary sites shared a common ancestor yet continued to evolve independently. In 53% of cases, we found clinically actionable driver alterations in the brain metastases that were not detectable in the matched primary-tumor sample. In contrast, spatially and temporally separated brain metastasis sites were more genetically homogenous and shared nearly all driver alterations detected. Extracranial metastases and regional lymph nodes were highly divergent from brain metastases. Several clinically actionable pathways were enriched in brain metastases. Conclusions: These observations demonstrate that brain metastasis tissue provides an opportunity to identify clinically important driver alterations that may be undetected in single samples of primary tumors, regional lymph nodes, or extracranial metastases. Genetic divergence between primary tumors and brain metastases may underlie some of the difficulties encountered with the combined treatment of systemic disease and brain metastases in patients. When clinically feasible, genomic characterization of brain metastasis tissue should be considered when selecting therapeutic agents for patients with brain metastases. Citation Format: Priscilla K. Brastianos, Scott L. Carter, Sandro Santagata, Daniel Cahill, Amaro Taylor-Weiner, Robert T. Jones, Eliezer Van Allen, Peleg Horowitz, Keith L. Ligon, William T. Curry, Ian F. Dunn, Paul Van Hummelen, Matthew Meyerson, Levi Garraway, Josep Tabernero, Joan Seoane, Stacey Gabriel, Eric S. Lander, Rameen Beroukhim, Tracy T. Batchelor, Jose Baselga, David N. Louis, William C. Hahn, Gad Getz. Genomic characterization of brain metastases reveals branched evolution and metastasis-specific mutations. [abstract]. In: Proceedings of the AACR Special Conference: Advances in Brain Cancer Research; May 27-30, 2015; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2015;75(23 Suppl):Abstract nr PR04.

RNA sequencing (RNASeq) provides the ability to comprehensively assay the transcriptome in a high-throughput manner. Current there are a variety of library preparation methodologies for measuring and sequencing the transcriptome depending on (a) the sample source and (b) outcomes of interest. Beyond protocol selection, the requisite computational tools and resources are significant considerations in processing, analyzing and reporting the experimental results. While there are many resources readily available to effectively perform RNA-seq experiments, optimal protocols and analysis tools for the cancer domain remain to be developed. We have developed and characterized a set of protocols and analysis procedures that comprise an RNA-seq pipeline that can effectively be used in a cancer research setting. The analysis pipeline consists of a sequence of functions and tools to process and clean the raw data, generate quality control and summary metrics, and perform secondary analyses that include expression quantification, fusion detection and somatic mutation calling. We applied this pipeline to three different RNAseq strategies (whole-transcriptome, exome, and targeted RNA-seq) and performed an in-depth comparative analysis to investigate the implications of the choice of strategy on the downstream analysis and results. More specifically, we investigated the impact of library preparation methods on the dynamic range and expression profiles, variant calling and fusion detection. While the data indicated that capture-based protocols provided efficient methods for sampling transcripts as compared to whole-transcriptome RNA-seq, there are considerations in its use, particularly for duplicate reads and uncaptured transcripts. We illustrate the implications of these issues on downstream analysis, such as somatic mutation and fusion calling and differential expression. In summary, we have described a RNA-seq analysis platform that provides a varied set of library preparations and analytical components for large-scale clinical or research transcriptomics. Our analysis has characterized the technical features of the different library preparations, providing a necessary understanding of the costs and benefits of each method and the potential effects on the downstream analyses. Citation Format: Ryan P. Abo, Ling Lin, Samuel S. Hunter, Deniz N. Dolcen, Rachel R. Paquette, Angelica Laing, Luc de Waal, Aaron R. Thorner, Matthew D. Ducar, Liuda Ziaugra, William C. Hahn, Matthew L. Meyerson, Laura E. MacConaill, Paul Van Hummelen. Comparative analysis of RNA sequencing methods for characterization of cancer transcriptomics. [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 4867. doi:10.1158/1538-7445.AM2015-4867

Background: Brain metastases represent an unmet need in current oncologic care. Approximately 8-10% of cancer patients will develop brain metastases, and more than half of these patients will pass away within a few months of their diagnosis. We have a limited understanding of how brain metastases genetically evolve from their primary tumors. Our objectives were to (1) elucidate the genomic evolutionary patterns leading to brain metastases (2) identify whether brain metastases harbor clinically significant genetic differences compared to their primary tumors and other extracranial metastatic sites, and (3) examine the extent of genetic heterogeneity across regionally separated and anatomically distinct sites of brain metastasis. Methods: We subjected 104 matched primary tumor biopsies, brain metastases, and normal tissue to whole exome sequencing, including 20 cases with regionally and anatomically separated brain metastasis sites, regional lymph nodes, and distal extracranial metastases. We performed an integrative analysis of somatic single nucleotide variants and copy-number alterations to reconstruct phylogenetic trees relating the subclones from each patient. We analyzed evolutionary relationships between related cancer samples and annotated phylogenetic trees with clinically significant genetic alterations. Results: Every brain metastasis displayed branched evolution: the brain metastasis and primary tumor shared a common ancestor yet both the primary tumor and brain metastasis continued to evolve independently. We found novel clinically actionable genetic alterations that were exclusive to brain metastases in 56% of cases. The brain metastases were also enriched for several pathways, some pathways specific to a particular histology. Distal extracranial metastases and regional lymph nodes were highly divergent from brain metastases, and in no cases, did we observe an extracranial site that closely resembled the brain metastasis. In contrast, regionally and anatomically separated brain metastasis sites were genetically homogenous and shared nearly all genetic alterations detected. Conclusions: Brain metastases are genetically divergent from primary tumors. Clinically, these observations demonstrate that biopsies of primary tumors fail to capture the heterogeneity with patients with brain metastases, potentially missing clinically actionable mutations in these life-threatening metastases. Notably, regional lymph nodes and distal extracranial metastases were not reliable genetic surrogates for brain metastases. When clinically feasible, characterization of even a single brain metastasis lesion is superior to that of a primary or lymph node biopsy for selection of a targeted therapeutic agent. Citation Format: Priscilla K. Brastianos, Scott L. Carter, Sandro Santagata, Amaro Taylor-Weiner, Robert T. Jones, Eli Van Allen, Keith L. Ligon, Josep Tabernero, Joan Seoane, Elena Martinez-Saez, Daniel Cahill, William T. Curry, Ian F. Dunn, Sun Ha Paek, Paul Van Hummelen, Aaron R. Thorner, Bruce E. Johnson, Nancy U. Lin, Toni K. Choueiri, Michael S. Rabin, Rameen Beroukhim, Anat Stemmer-Rachamimov, Matthew Meyerson, Levi Garraway, Tracy Batchelor, Jose Baselga, David N. Louis, William C. Hahn, Gad Getz. Genomic characterization of brain metastases reveals divergent evolution and metastasis specific mutations. [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 4727. doi:10.1158/1538-7445.AM2015-4727

Identification of copy number alterations (CNAs) in tumors can assist in determining diagnosis, prognosis, or predicting response to therapy in certain cancer types. Many CNA detection tools have been developed for whole genome sequencing (WGS). These algorithms typically “call” alterations by using significant deviations of the log2 ratio of tumor to normal coverage from the expected diploid log2 ratio. Several factors, however, including reporting pertinent genomic information, hardware and software limitations, and cost, have impeded the routine application of WGS in a clinical setting. Targeted sequencing - hybrid capture and massively parallel sequencing of defined regions (targets) of the genome can overcome some of these challenges and can specifically interrogate, for example, known cancer genes, thus maximizing the “actionability” or utility of the information generated. Though more suited for large-scale clinical testing, targeted sequencing of smaller gene panels with specific clinical reference presents some significant bioinformatic challenges that hinder the application of pre-existing CNA detection tools, resulting in a heavy dependence on manual review in the clinic. In addition, varying quality of DNA, minor variations in laboratory protocols, along with batch effects, are known to confound CNA profiles and are exasperated by the decreased number of sequenced regions of targeted gene panels. These variables alter the log2 ratio, and thus, our ability to call CNAs suffers. We note, that using matched normals may mitigate some deleterious effects, however, to minimize clinical cost and throughput, we tackle these challenges with unmatched tumor-normal pairs. Similar to the closest normal approach used in SNP array analyses, we rely on the assumption that some “normal” (non-tumor) samples experience similar experimental variability that affects the tumor samples. Minimizing maximally informative cost-functions (incorporating both global and local CNA variation), we iteratively investigate different combinations of normal samples which best resemble a given cancer sample. The new set of normals which minimize the cost-function, defines a panel of normals (PON) whose target medians are used to generate the log2 ratio. Along with batch-effect reduction techniques, we then call CNAs with statistically significant thresholds at the target-level. Interestingly, we find that approximately 15 normals are typically needed to minimize CNA profile variation, as opposed to our typical approach of using all available normals (∼60). In addition, we demonstrate that our procedure can improve our sensitivity and specificity from 58±1% to 81±2% and 94±1% to 99.5±0.1%, respectively, on simulated data. We further compared our approach with other published algorithms and found a significant improvement in performance, reducing the time required to manually review CNAs and interpret copy number profiles in the clinic. Citation Format: Bernard Fendler, Ryan Abo, Samuel Hunter, Matthew Ducar, Elizabeth Garcia, Paul Van Hummelen, Neal Lindeman, Laura MacConaill. Identifying copy number alterations from targeted sequencing data. [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 4850. doi:10.1158/1538-7445.AM2015-4850

Identification of structural gene rearrangements is vital for cancer patients as these events can provide definitive diagnoses, prognostic value, and influence the course of treatment. While FISH, karyotype analysis and aCGH array have traditionally been used to identify and confirm the presence of structural variants, the advent of next generation sequencing has enabled genetic testing including detection of multiple structural variants (SVs) from genomic DNA. To this end, we have developed and validated Oncopanel, a cancer-specific targeted next generation sequencing (NGS) assay designed to detect SNVs, indels, and copy number alterations across 300 genes, and 35 clinically actionable or informative SVs. Each rearrangement was specifically targeted by baiting genomic locations frequently identified to contain breakpoints reported in the literature and publicly available databases. Using BreaKmer, an internally developed SV detection tool (Nucleic Acids Res. 2014 Nov 26, doi: 10.1093/nar/gku1211), rearrangements, including the exact breakpoint coordinates and the genes involved in or adjacent to the breakpoint(s), were identified. Here we examine the utility of Oncopanel using genomic DNA to identify structural variants. We report the results of 3,291 cancer patients tested in our personalized cancer medicine program (Profile), a joint venture between Dana-Farber Cancer Institute, Brigham and Women's Hospital, and Boston Children's Hospital. As compared to conventional cytogenetics, FISH analysis, and molecular detection by PCR methods, Oncopanel's overall sensitivity and specificity for SVs was 81.4% and 100%, respectively. Most discordant results were identified in (1) tumors with SVs involving the IGH enhancer regions (60% of discordant results), or (2) in samples with < 20% tumor (25% of discordant results). Several SVs involving the IGH enhancer regions were missed likely due to lack of Oncopanel coverage. Oncopanel was designed to target a finite sequence pool, but due to IGH enhancer region's large size (1.2Mb), only a small portion of this region was specifically interrogated. Inclusion of all possible IGH enhancer sequences would have hampered the effectiveness of SNV, indel and copy number alteration detection for other cancer critical genes. Discrepant Oncopanel and cytogenetic results were also observed in samples with low tumor purity likely due to masking of variant sequences by stromal contamination. In conclusion, we find that Oncopanel has utility to detect structural variants with a sensitivity of 92%, barring detection of rearrangements involving IGH, and a specificity of 100%. Based on the baiting strategy, detection of many rearrangements can also be interrogated in parallel with SNV, indel and CNV detection resulting in reduced sample input requirements and the inclusion of precise information regarding the breakpoints and the class of rearrangement identified. Citation Format: Elizabeth P. Garcia, Azra H. Ligon, Ryan P. Abo, Paola S. Dal Cin, Stanislawa Weremowicz, Priyanka Shivdasani, Phani K. Davineni, Dimity L. Zepf, Matthew D. Ducar, Paul Van Hummelen, Yonghui Jia, Frank C. Kuo, Lynette M. Sholl, Laura E. MacConaill, Neal I. Lindeman. Detection of gene rearrangements using OncoPanel: a targeted next-generation sequencing assay. [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 2991. doi:10.1158/1538-7445.AM2015-2991

Background: The overall death rate from lung cancer is higher in African-Americans (AA) compared to Caucasians (CAU). Understanding differences in the prevalence and type of somatic alterations between races may illuminate differences in prognosis and lead to the reduction of outcome disparities by more precisely targeting patients’ treatment. Methods: Formalin-fixed paraffin embedded tumor samples were collected from Baptist Cancer Center, Memphis, TN. DNA was extracted and sonicated to 250 bp following Covaris FFPE DNA Extraction & Purification protocol and further purified using Agencourt AMPure XP beads. The OncoPanel_v2 target enrichment panel (Agilent SureSelect) was used for hybrid capture of 502 cancer-related genes. Samples were pooled and sequenced on an Illumina HiSeq2500 to a mean depth of coverage of 210x. Tumors with >30x sequencing depth over >80% of targeted bases were considered for analysis. MuTect and SomaticIndelDetector were used to identify somatic single nucleotide variants (SNVs) and short insertions or deletions (indels), respectively. As matched normal DNA was not available for these samples, analysis was limited to variants previously observed in tumors as described in the Catalogue of Somatic Mutations in Cancer (COSMIC) database, and variants were excluded if found in the germline dataset of the Exome Sequencing Project (ESP). Results: 510 tumor specimens from 242 Black and 268 White patients (with self-reported race) were analyzed including 320 adenocarcinomas and 142 squamous cell carcinomas. Pathological classification was independently reviewed and confirmed for 374 of 472 cases for an overall concordance rate of 79%. Using principle component analysis (PCA) on germline SNVs, we observed that the biological ancestry was different than the self-reported race for 1.5% of patients. Mutational frequencies for genes with known roles in adenocarcinoma such as KRAS and EGFR were not significantly different between tumors from Black and White patients (Fisher's exact p-value > 0.05). Amplification rates for NKX2-1, MET, MDM2, and MYC and homozygous deletion rates for CDKN2A were also not significantly different between populations. Translocations involving ALK and ROS1 were detected in tumors from Black patients demonstrating that these events are present in both populations. Similarly, mutational frequencies for genes such as PIK3CA and PTEN were not significantly different across populations in squamous cell carcinomas. Conclusions: These results demonstrate that lung cancers from Black patients are more similar to Whites than East Asians with respect to genes such as EGFR, and suggest that clinical trials of targeted therapies could significantly benefit patients in both populations. Citation Format: Joshua Campbell, Christopher Lathan, Lynette Sholl, Matthew Ducar, Mikenah Vega, Ling Lin, Aaron Thorner, Nick Faris, Paul van Hummelen, Raymond Osarogiagbon, Matthew Meyerson, Laura MacConaill. Comparing the mutational landscape of African American and Caucasian lung cancers. [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 3886. doi:10.1158/1538-7445.AM2015-3886

RNA sequencing (RNA-seq) is a transcriptome profiling technology that provides multiple levels of insight into the genome. In addition to expression levels (transcript abundance), it generates endpoints such as alternative splicing, somatic mutations and rearrangements, which may have functional consequences in cancer. Although somatic mutations are generally identified by DNA sequencing, RNA-seq has the advantage of detecting allele-specific expression affecting a variant allele, as well as functional chimeric transcripts that result from structural rearrangements. Compared to microarray technologies, RNA-seq can provide additional information about novel transcripts. Due to the complexity of the human transcriptome and the variability of gene abundance, the cost of whole transcriptome sequencing to achieve sufficient coverage to detect these types of alterations remains high. To explore the feasibility of a more cost-effective method, we compared the performance of three different RNA-seq methods: whole-transcriptome-, exome-, and targeted RNA-seq, using RNA derived from cancer cell lines and Formaldehyde Fixed-Paraffin Embedded (FFPE) samples. For whole-transcriptome preparation, we used the Illumina TruSeq Stranded mRNA and total RNA kits for cell line and FFPE samples, respectively. Exome-RNAseq was performed using the Illumina Access kit. The libraries from whole-transcriptome RNAseq were subjected to hybridization capture using OncoPanel-an Agilent SureSelect baitset of 500 cancer-related genes. Compared to whole-transcriptome, exome- and targeted-RNA-seq demonstrated (1) higher coding exon coverage and multiplexing capability; (2) reduced rRNA composition to 1%; (3) comparable gene abundance information and (4) over 90% of reads aligned to coding exon regions in FFPE samples, compare to ∼30% when using whole transcriptome method. In conclusion, we demonstrated that exome- and targeted RNA-seq provide a cost-effective way to analyze a subset of the transcriptome. Furthermore, targeted RNA-seq can be highly multiplexed and is therefore amenable for large-scale tumor profiling in clinical or research settings. Citation Format: Ling Lin, Ryan Abo, Deniz Dolcen, Rachel Paquette, Angelica Laing, Luc de Waal, Aaron Thorner, Matthew Ducar, Liuda Ziaugra, Bruce Wollison, Marc Breneiser, William Hahn, Matthew Meyerson, Paul Van Hummelen, Laura MacConaill. Targeted RNA sequencing improves transcript analysis in cancer samples. [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 1115. doi:10.1158/1538-7445.AM2015-1115

The objective of this research was to determine whether Novantrone, Oncovin, Velban, and Prednisone (NOVP) combination chemotherapy for Hodgkin's disease increases the frequencies of the specific types of aneuploid sperm that might elevate the risk of fathering a child with one of the major clinical aneuploidy syndromes, i.e., Down (disomy 21 sperm), Edward (disomy 18 sperm), Turner (nullisomy sex sperm), XYY (disomy Y sperm), triple X (disomy X sperm), or Klinefelter (XY sperm). A four-chromosome multicolor sperm fluorescence in-situ hybridization assay that simultaneously evaluates chromosomes 18, 21, X, and Y was applied to semen provided by four healthy men and to repeated samples of eight Hodgkin's disease patients before treatment, 35-50 days after treatment to examine the effects of treatment on male meiotic cells, and 1-2 years after treatment to measure the persistence of damage. There were chromosome-specific variations in baseline frequencies and significant inductions of all of the detectable types of sperm aneuploidies: XY sperm (14-fold increase), disomy 18 (7-fold), nullisomy sex (3-fold), disomy 21 (3-fold), and disomy X and Y (approximately 2-fold each). Disomy 21 was about twice as frequent as disomy 18, and neither showed a preferential segregation with a sex chromosome. Extrapolating across the genome, approximately 18% of sperm carried a numerical abnormality after NOVP treatment of meiotic cells. Induced effects did not persist to 1-2 years after treatment, suggesting that persistent spermatogonial stem cells were not sensitive to NOVP. These findings establish the hypothesis that conception shortly after certain chemotherapies can transiently increase the risks of fathering aneuploid pregnancies that terminate during development or result in the birth of children with major human aneuploidy syndromes.

Background: Brain metastases are the most frequently occurring intracranial tumors in adults. Median survival after the diagnosis of a brain metastasis is in the order of a few months. Despite its large burden of disease and devastating clinical sequelae, we continue to have a limited understanding of how brain metastases evolve from their primary tumor. Our objectives were to (1) elucidate the evolutionary patterns leading the brain metastases and (2) identify whether brain metastases are genetically distinct from their matched primary tumors. Materials and Methods: We subjected 101 trios consisting of primary tumor, brain metastasis, and matched normal tissue to whole exome sequencing (WES). To analyze the data, we developed novel computational tools to perform an integrative analysis of somatic single nucleotide variants (SSNVs) and somatic copy-number alterations (SCNAs). This analysis allowed us to estimate the clonal architecture of the primary and metastatic samples from each patient, and to reconstruct a phylogenetic tree relating all of the subclones.Results: Every metastasis developed from a single clone, consistent with a single cell of origin. We did not detect evidence of self-seeding or a multiclonal origin of metastasis. In all cases, we observed a sibling or a branched evolutionary relationship; the brain metastasis and primary tumor share a common ancestor, but there was continued evolution in the primary tumor reflected by fully clonal mutations in the primary biopsy that were not present in the metastasis. When we average over all the phylogenetic trees, 61% of mutations are present in the common ancestor, 24% are unique to the metastasis and 15% are unique to the primary tumor. Subclonal mutations in the metastasis by definition occurred within the brain; these mutations displayed different mutational signatures than those acquired in the primary tumor. These contrasts were most pronounced in cases of lung cancer or melanoma, with tobacco and UV signatures prominent in these primaries and nearly absent from the mutations acquired after metastasis. In order to understand the molecular drivers of clonal evolution and metastasis in our data, we annotated each subclone with driver mutations identified using large numbers of cancer samples analyzed by the cancer genome atlas (TCGA) consortium. This produced a detailed portrait of each patient's cancer, with nearly node in each phylogenetic tree associated with at least one known driver. We found novel drivers, many of which are known actionable targets, in the clonal and subclonal populations within the brain metastases that were not present in the primary tumor. This suggests ongoing evolution within the brain. Similarly, novel subclonal and clonal drivers were detected in the biopsy of the primary tumor that were not present in the metastasis. The brain metastases were enriched for several pathways when compared to their matched primary tumors, some pathways specific to a particular histologic subtype.Conclusions: In this study, we report the latest results of the largest massively parallel sequencing study to date of matched brain metastases and primary tumors. We used intratumoral heterogeneity estimates to elucidate the evolutionary patterns observed in the process of metastasis. This study shifts our understanding of the metastasis paradigm and sheds light on the evolutionary and molecular mechanisms that are critical for brain metastasis. Our data suggests that single biopsies do not capture the heterogeneity within patients. Assessment of the subclonal phylogenetic architecture of primaries and their metastases should be considered when selecting targeted agents for patients with brain metastases. Citation Format: Priscilla K. Brastianos, Scott L. Carter, Sandro Santagata, Amaro Taylor-Weiner, Robert T. Jones, Peleg M. Horowitz, Keith L. Ligon, Joan Seaone, Elena Martinez-Saez, Josep Tabernero, Daniel P. Cahill, Sun-Ha Paek, Ian F. Dunn, Bruce E. Johnson, Toni K. Choueiri, Michael S. Rabin, Eric P. Winer, Nancy U. Lin, Paul Van Hummelen, Anat Stemmer-Rachamimov, Rameen Beroukhim, David L. Louis, Tracy T. Batchelor, Jose Baselga, Gad Getz, William C. Hahn. Genomic characterization of 101 brain metastases and paired primary tumors reveals patterns of clonal evolution and selection of driver mutations. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr NG03. doi:10.1158/1538-7445.AM2014-NG03

Targeted next-generation sequencing (NGS) to capture genes or regions of interest has proven to be a cost-effective alternative to whole genome sequencing (WGS), particularly for cancer research and clinical cancer care. In this context, biologically or clinically relevant selected genes/regions are sequenced to several hundred-fold coverage. However, current algorithms and tools for detecting large structural variants (SVs), such as translocations, fail to achieve either high specificity or sensitivity, due to the fact that most available methods were designed for WGS data, and thus do not take advantage of the reduced size and higher coverage of targeted sequencing to improve SV calling. We developed a novel method, BreaKmer, to detect indels, rearrangements, and translocations from single sample targeted genomic reads. The algorithm extracts mis-mapped single or paired-end NGS reads. From these reads, hypothesized to contain SV breakpoints, contigs are built with a kmer strategy: reads are broken into k-length substrings or kmers, and those occurring in the reference are filtered. The remaining kmers represent sequences containing any sequence variant from the reference, ranging from single nucleotides to larger variants. Contigs are assembled from reads containing sample specific kmers. SVs are called based on alignment of the contigs to the reference sequence. With paired-end (PE) reads, discordantly mapped paired reads are extracted and coupled with SV calls that are made. To demonstrate BreaKmer, we analyzed NGS data from 166 samples enriched using 3 different capture panels (ranging from 305-504 genes). Our dataset contained 25 cancer specimens with known translocation events verified by orthogonal clinical methods and a negative control set of 141 ‘normal’ samples with no known SVs. The samples represented DNA extracted from FFPE, fresh frozen, blood and cell lines. Among 25 samples, 15 had additional replicate samples. Mean target coverage over all the samples was on average 150x. Specimens were barcoded at library preparation and pooled, followed by hybrid-capture targeting cancer genes and sequenced 2x100bp PE. All translocation events from the 25 test samples and their replicates (46) were detected by BreaKmer. An additional 19 translocations were detected among all the samples and their replicates, while no translocation calls were made for the 141 negative control samples. Our novel kmer strategy to detect SVs displayed high sensitivity and specificity. We reliably detected rearrangements of ALK, BCL2-IGH, BCR-ABL1 from lung adenocarcinoma, B-cell lymphoma, and chronic myeloid leukemia samples, respectively; indicating real clinical utility of this algorithm. In addition, our tool effectively detects other SV types - such as indels in FLT3 and KIT, among other genes. Our algorithm thus serves a pressing need for improved SV detection in targeted NGS data, particularly in precision cancer medicine. Citation Format: Ryan P. Abo, Elizabeth P. Garcia, Matthew Ducar, Ravali Adusumilli, Marc Breneiser, Vanessa Rojas-Rudilla, Lynette M. Sholl, Neal I. Lindeman, Matthew L. Meyerson, William C. Hahn, Paul Van Hummelen, Laura E. MacConaill. BreaKmer: Detection of structural rearrangements in targeted next-generation sequencing data using kmers. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 5321. doi:10.1158/1538-7445.AM2014-5321

Next-generation sequencing (NGS) is increasingly used in the laboratory and clinic to identify genomic factors contributing to tumorigenesis, such as somatic mutations, DNA insertions and deletions, transcriptome and epigenetic changes, and chromosomal abnormalities. Because tumor specimen tissues are often quite limited or are from rare and precious samples, it is necessary to prepare sequencing libraries from minimal amounts of DNA. Consequently, limited DNA input necessitates amplification during NGS library preparation. Therefore, it is of great importance to amplify the library in an efficient and uniform manner to achieve reproducible amplification of the library fragments and limit GC content bias. Variable GC-bias among libraries is of particular importance for copy number variant (CNV) detection based on sequence coverage. Here, we compare the TruSeq DNA Sample Preparation Kit (Illumina Inc., San Diego, CA) and the KAPA Library Preparation Kit (Kapa Biosystems, Wilmington, MA) to determine which technique achieves optimal library coverage. The major difference between the methods is the DNA polymerase used for library enrichment. The Kapa HiFi polymerase was engineered to have higher fidelity and increased DNA affinity relative to the Herculase II Fusion DNA Polymerase (Agilent Technologies, Inc., Santa Clara, CA) that we used for library enrichment with the TruSeq kit. Thus, the Kapa HiFi polymerase should more efficiently amplify targets across a larger GC content range and improve sequence coverage uniformity. Relative to TruSeq-prepared libraries, the Kapa-prepared libraries had a larger mean insert size, more uniform coverage distribution, and better coverage of GC-rich regions of the genome. This was observed for DNA extracted from blood, cell lines, and formalin-fixed paraffin embedded (FFPE) tumor samples. More experiments are underway to determine the effect of the improved recovery of high-GC content DNA fragments on CNV detection. Citation Format: Aaron R. Thorner, Ashwini Sunkavalli, Ling Lin, Robert T. Jones, Laura Schubert, Matthew D. Ducar, Ravali Adusumilli, Deniz Nesli Dolcen, Liuda Ziaugra, Jack Lepine, Laura E. MacConaill, William C. Hahn, Matthew Meyerson, Paul Van Hummelen. Reducing GC-bias and improving coverage distribution in Illumina sequencing using the Kapa Biosystems library construction method. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 3583. doi:10.1158/1538-7445.AM2014-3583

Formalin fixation followed by paraffin embedding (FFPE) is the most widely used method of tissue preservation and often used in the clinical setting for cancer specimens. However, this process of fixation has negative consequences on the quality of extracted DNA, and inevitably adversely affects downstream analyses such as next generation sequencing (NGS). DNA extraction from FFPE samples requires paraffin disassociation, tissue rehydration and de-crosslinking. Low efficiency in any step will yield low quantities of nucleic acids with incomplete reversal of formaldehyde crosslinks, and mixtures of single and double stranded DNA. In addition, low quality nucleic acids will often result in poor sequencing quality due to inefficient library preparation, lower number of reads with high quality, and variable recovery of sequence reads with high GC-content. Here we describe the use of Adaptive Focused Acoustics technology (Covaris Inc. 2013) to remove paraffin and break down tissue, and we compared this method to traditional xylene-based paraffin removal. We found that Covaris had superior performance overall: (1) DNA yields were 6 times higher, and less specimens failed extraction: this is critically important when dealing with small tumor biopsies. (2) DNA was more readily available for PCR amplification. (3) Mean coverage of DNA fragments with high GC content was higher and more reproducible. (4) Library complexity was higher. In addition, the Covaris protocol required less hands-on time in the lab, and eliminated the need for hazardous organic solvents for paraffin removal. In conclusion, the Covaris FFPE DNA extraction method is very efficient, enhances DNA availability, improves reproducibility and provides higher DNA quality for NGS. Citation Format: Ling Lin, Mikenah Vega, Robert T. Jones, Liuda Ziaugra, Deniz N. Dolcen, Ashwini Sunkavalli, Laura Schubert, Jack R. Lepine, Aaron R. Thorner, Matthew D. Ducar, William C. Hahn, Matthew L. Meyerson, Laura E. MacConaill, Paul Van Hummelen. Improved FFPE DNA extraction for next generation sequencing using adaptive focused acoustics technology. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4286. doi:10.1158/1538-7445.AM2014-4286

Meningiomas are the most common primary brain tumor in the US. While many Grade I meningiomas are cured by surgery, 20% recur and are minimally responsive to systemic therapy. Meningiomas of higher grades are particularly challenging to treat and are associated with a poor prognosis. Though the tumor suppressor gene NF2 has been linked to meningiomagenesis, the spectrum of genetic changes in meningiomas remains poorly understood, particularly in the large subset of tumors without NF2 alterations. We performed whole-exome or whole-genome sequencing on 17 meningiomas and focused sequencing using a 645 gene-set on an additional 48 tumors. The majority of meningiomas exhibited simple genomes, with fewer copy-number alterations, mutations, and rearrangements compared to other adult tumors. However, several meningiomas harbored more complex patterns of copy-number changes and rearrangements including one tumor with chromothripsis. We confirmed focal NF2 inactivation in 43% of tumors. A subset of NF2-wildtype meningiomas harbored recurrent oncogenic mutations in AKT1 and the Hedgehog pathway regulator SMO. Tumors with these mutations varied histologically from their NF2-mutated counterparts, and exhibited immunohistochemical evidence of pathway activation. These results begin to define the spectrum of genetic alterations in meningiomas. Our findings of known driver oncogenic mutations including AKT1 and SMO have immediate therapeutic potential. Citation Format: Priscilla Brastianos, Peleg M. Horowitz, Sandro Santagata, Robert T. Jones, Aaron McKenna, Gad Getz, Keith Ligon, Paul Van Hummelen, Matt Ducar, Alina Raza, Ashwini Sunkavalli, Laura MacConaill, Anat Stemmer-Rachamimov, David N. Louis, William C. Hahn, Ian Dunn, Rameen Beroukhim. Genomic sequencing of meningiomas reveals oncogenic SMO and AKT1 mutations. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 803. doi:10.1158/1538-7445.AM2013-803

The ability to harness the power of next generation sequencing for characterizing the cancer genome would be extremely valuable to the cancer research community and ultimately improve diagnosis and individualized therapy. We present a targeted approach using massively parallel sequencing to simultaneously detect mutations, translocations and copy-number variations in archived clinical tumor specimen. Targeted sequencing was achieved by designing RNA baits to capture the exons of 504 genes with relevance to cancer. The bait set was augmented with specific intronic sequences to detect translocations often involved in cancer. The overall performance of the hybrid capture was improved by optimizing the tiling strategy of the baits. These improvements significantly shifted fragments with low or no-coverage closer to the overall mean, resulting in a more uniform coverage distribution. To validate the targeted gene panel we sequenced tumor samples harboring mutations, translocations and copy-number alterations that were previously identified by clinically approved assays. All the known alterations were confirmed with additional potentially actionable mutations. The ultimate goal of the targeted panel is to screen cancer patients quickly and economically allowing simultaneous detection of all common genomic alterations in the cancer genome. Citation Format: Paul Van Hummelen, Matthew Ducar, Robert T. Jones, Alina Raza, Ashwini Sunkavalli, Megan Hanna, Adri Mills, Ravali Adusumilli, Prateek Kumar, Laura Schubert, Marc Breneiser, Anna C. Cooley, Elizabeth Garcia, Lynette M. Scholl, Neal I. Lindeman, Nikhil Wagle, Levi Garraway, Kristian Cibulskis, Scott L. Carter, Michael Lawrence, Gad Getz, Matthew L. Meyerson, William C. Hahn, Laura E. MacConaill. Targeted sequencing to detect somatic mutations, translocations and copy-number variation in human tumors simultaneously. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 819. doi:10.1158/1538-7445.AM2013-819

Detection of critical cancer gene mutations in tumor samples may result in improved care, treatment and outcomes for patients. It has become clear that for specific cancer types there is a high association between mutations in a defined set of oncogenes or tumor suppressors some of which are actionable. While it has become common practice to screen for specific mutations in specific cancer types, to date broad mutation screening has not been practical. We report here the first large scale screening initiative between the Dana Farber Cancer Institute and Brigham and Women's Hospital that allows adult cancer patients the option to have their tumor tissue screened for genetic mutations implicated in cancer. OncoMap is a high throughput mutation profiling platform developed to detect 471 genetic mutations across 41 known oncogenes and tumor suppressor genes. The goal is to utilize OncoMap to identify genetic mutations in cancerous tissue across all tumor and tissue types. Because there is no bias as to which cancers are screened for specific mutations our aim is to identify new trends and associations between cancer types, mutational status and patient outcome. OncoMap has been analytically validated under CLIA guidelines as recommended by the Clinical Laboratory Standards Institute. Our findings show that OncoMap detects somatic mutations that are present at an allele frequency of 7.5% with 95% confidence. Precision and reproducibility studies were performed using 30 distinct formalin-fixed paraffin embedded tumor samples from various tissue types known to harbor somatic mutations in KRAS, BRAF or exon 19 deletions in EGFR. Further concordance studies demonstrate that OncoMap shows ninety-six percent sensitivity with respect to detection of codon 12 and 13 mutations in KRAS, detection of BRAF V600E mutations and JAK2 V617F mutations. Using the OncoMap cancer genetic profiling platform we anticipate that ∼10,000 individual tumors from adult cancer patients will be screened on a yearly basis enabling the identification of newly discovered associations between cancer tissues, their mutational status and patient outcome. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 5058. doi:1538-7445.AM2012-5058

Although advances in technology and the biological understanding of cancer pathogenesis have brought the goal of personalized cancer medicine closer, important challenges remain in several areas. One such challenge is the reliable and sensitive detection of somatic mutations in clinical tumor specimens. Somatic mutation detection in clinical tumor specimens is confounded by biological variables such as the percentage of tumor cells in a specimen, stromal contamination, sequence quality (particularly in DNA derived from formalin-fixed clinical samples), tumor heterogeneity and ploidy considerations. These and other experimental factors can lead to allele signal ratios very different from the 1:0, 1:1 or 0:1 ratio expected for a classically-defined SNP. Germline genotyping using the Sequenom MALDI-TOF mass spectrometry platform has been used to reliably detect single-nucleotide polymorphisms (SNPs) in a wide variety of samples. The automated genotype calling algorithm included with the Sequenom platform falters, however, when the ratio of allele peak signals for a given assay in a specific sample do not approximate the native algorithm's expectations of frequency, confounding accurate genotype calling by the system. We report here a cancer-specific algorithm for sensitive genotype calling of somatic mutations, based on an independent clustering of raw allele peak data for each assay run, and the identification and stratification of outliers from an empirical background (‘wildtype’) cluster of genotypes. Outliers from this cluster are then evaluated using a number of criteria to determine the quality and performance of the overall assay run, as well as the individual call; these factors include the relative signal-to-noise ratio (SNR) of each allele peak, the ratio of variant allele peak height to peak area, the percentage of signal lost in unextended probe (UEP) and relative distance of a putative mutation call from the wildtype cluster In a scatter plot of peak area values. Based on these criteria, a confidence value is assigned to each candidate outlier, allowing for stratification and prioritization of these candidates for validation. We applied the algorithm to samples run against the our OncoMap genotyping panel, comprised of 460 assays interrogating mutations in 33 cancer-associated genes. Candidate mutations produced by this algorithm correlate strongly with those called by users manually, and use of the algorithm increases the sensitivity and reliability of the OncoMap platform to detect somatic mutations in clinical tumor samples, particularly in FFPE-derived material. The algorithm code and documentation will be made publicly available in a form compatible with most Sequenom genotyping platform installations. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 3965. doi:1538-7445.AM2012-3965