470 Background: A comprehensive molecular characterization of primary colorectal cancers (CRC) was recently reported. Less is known about mutation patterns in CRC metastases and association with survival. Our sequencing analysis focused on CRC liver metastases with RAS/RAF mutations, representing a patient population with limited therapeutic options. Methods: DNA was extracted from formalin-fixed paraffin-embedded CRC liver metastases. Fifty tumors found by Sequenom MassARRAY to harbor KRAS, NRAS or BRAF mutations underwent next generation sequencing on the Ion AmpliSeq Comprehensive Cancer Panel of 409 genes. Co-investigators were blinded to Sequenom mutations identified at UCSF. Variants called by Strelka and VarScan were extensively filtered to control the False Positive Rate and find mutations occurring with > 5-10% variant allele frequency compared to normals. The dataset was evaluated for significant co-mutations, biclustering, and population probabilities of mutations. Results: Following sequencing, 37,744 variants were called in 409 genes with a median coverage depth of 1053x. After filtering to minimize false positives, 2335 variants in 315 genes remained. ARID1A and PIK3R1 were the most significantly associated co-mutation pair, P < 3.5e-5. Biclustering showed no stratification of patients; genes stratified only by mutation frequency. Further filtering yielded 1,186 variants present at < 1% allele frequency within 1,000 Genomes, of which 131 variants in 24 genes are referenced in the Catalog of Somatic Mutations in Cancer. In addition to anticipated mutations in mismatch repair genes and the RTK/RAS/PI3K, Wnt, TP53, and TGF beta pathways, infrequent mutations were found in Akt1, mTOR, MET and PPP2R1a. After APC, TP53 was the most commonly mutated gene, in 44% of the tumors (95% cl: 31.1% - 57.8%). Survival was similar with mutation of RAS/RAF plus either TP53 or PIK3CA. Conclusions: Next generation sequencing was used to characterize co-variants in RAS/RAF mutated CRC liver metastases. The complexity of our results is consistent with the clinical observation that targeting RAS/RAF mutated metastatic CRC is a formidable challenge. These analyses may nonetheless inform the design of future clinical trials.
Pool‐hmm: a Python program for estimating the allele frequency spectrum and detecting selective sweeps from next generation sequencing of pooled samples
