Detection of mutational signatures in advanced cancer patients using clinical-grade specimens.

3137 Background: The underlying extrinsic and/or intrinsic mutational processes underlying tumor development lead to genome-wide “passenger” alterations that can be identified as mutational signatures. Mutational signatures are known to be both prognostic and predictive in certain clinical contexts. Whole genome sequencing (WGS) based approaches are optimal for detection of mutational signatures but to date studies using formalin fixed paraffin embedded (FFPE) clinical grade specimens are limited. Methods: Archival DNA specimens from tumors of United States Veterans that underwent clinical sequencing in the VA National Precision Oncology Program were subjected to 30X WGS. Somatic alterations were identified by a tumor-only variant calling method employing filtering of variants found in the dataset at >10% frequency and putative germline variants found in multiple databases and germline WGS from the Million Veteran Program (MVP). Mutational signatures were identified using deconstructSigs. Results: WGS was performed on 134 tumor samples with matched germline WGS from MVP; variants were called using the tumor-only variant calling pipeline in addition to tumor-normal calling. Tumor-only WGS derived tumor mutational burden (TMB) was highly correlated with TMB from clinical targeted sequencing (White, R2=0.48; Black, R2=0.63; Other, R2=0.85) and with TMB from tumor-normal variant calling (White, R2=0.96; Black, R2=0.63; Other, R2=0.85). We used our tumor-only variant calling pipeline on WGS from 397 tumors. The median (+SD) number of mutations per tumor varied per tumor type, with the lowest number in head and neck squamous cell carcinomas (12445+4186 mutations) and highest in melanoma (119089+259854 mutations). Overall, 66 single base substitution (SBS) mutational signatures were identified. The SBS1 clock signature was found in 339 of 397 tumors. The SBS4 smoking mutational signature was found in 100% of small cell lung cancers (n=5), 76% of lung squamous cell cancers (n=38), 50% of lung adenocarcinomas (n=50), 36% of cancers of unknown primary (n=14), 29% of bladder cancers (n=7), and 14% of HNSC (n=36), but was not found in any other cancer type. The SBS7a UV mutational signature was found in 80% of melanomas (n=10) and 7% of cancers of unknown primary, but not in any other cancer type. Finally, the SBS6 and SBS15 mismatch repair mutational signatures were found in 10 and 12 of 16 MSI-high tumors, respectively. Conclusions: We have established a pipeline for evaluation of mutational signatures from tumor-only WGS of clinical grade FFPE tumors. Preliminary analysis shows identification of expected mutational signatures. Work is ongoing to determine whether ancestry matched panel of normals improves signature identification and to determine the prognostic and/or predictive ability of mutational signatures in a variety of clinical contexts which could greatly expand the clinical utility of WGS in advanced cancers.