Integrating molecular profiling into cancer treatment decision making: Experience with over 35,000 cases.

11001 Background: Molecular profiling of both common and rare cancer types provides for the identification of actionable targets for chemotherapy with many unexpected associations. Methods: Caris Life Sciences database of >35,000 profiled cancers was reviewed for well-established driver gene mutations and copy number alterations, and protein expression patterns that are relevant for selection of targeted therapy. Based on the published literature, these tumor characteristics were then associated with potential benefit or no benefit to the specific therapeutic agents. All relevant published studies were evaluated using the USPSTF grading scheme for study design and validity. Assay methodologies included sequencing (Sanger, pyrosequencing), PCR, FISH, CISH, and immunohistochemistry. Results: All common malignancies (10 most common cancer types in men and women) and 10 rare cancer types were well represented (minimum of 100 cases in each individual cancer type). Well established driver mutations and protein expression in common cancers were all identified with expected frequencies (e.g. HER2 amplification in breast, PIK3CA mutations in ER+ breast cancer, EGFR mutations in NSCLC, etc.). Importantly, unexpected new and potentially actionable targets were identified in common (e.g., 6.7% HER2 amplification in NSCLC, 1.6% KRAS mutation in prostatic adenocarcinoma) and rare cancers (e.g., 8.3% ALK alteration in soft tissue sarcomas, 10.5% c-MET and 26.4% EGFR gene amplification in melanomas, 16.3% KRAS mutation in cholangiocarcinomas, 10% AR expression in STS), as well in cancers of unknown primary site (approximately 4% of all tested cases). Conclusions: This review of the large referral cancer profiling database provided an unparalleled insight in the distribution of common and rare genetic and protein alterations with direct and potential treatment implications. Numerous targets were discovered that had a potential to be treated by the conventional chemotherapy as well as targeted therapy not usually considered for the cancer type. Comparison between an individual patient tumor profile and database for the matched cancer type provides additional level of support for targeted treatment choices.