Targeted exome sequencing of 69 human melanoma brain metastases

In-depth information regarding the DFCI OncoPanel sequencing panel has been described previously. Briefly, sequencing is performed using an Illumina HiSeq 2500 system (RRID:SCR_016383) with 2×100 paired-end reads. Samples must meet an average 50X coverage and minimum of 30X coverage for 80% of targets for analysis. For all samples, a board-certified molecular pathologist evaluates alterations and writes interpretations of the results. The three generations of OncoPanel use Mutect (RRID:SCR_000559) and GATK (RRID:SCR_001876) to interrogate possible alterations in the complete exonic DNA sequences of 275, 309, and 447 cancer-related genes, respectively, including substitutions, insertions, and deletions. Because tumor tissues are tested without a paired normal from individual patients, additional informatics steps are taken to identify common single nucleotide polymorphisms (SNPs). Any SNP present at >0.1% in Exome Variant Server, NHLBI GO Exome Sequencing Project (ESP) (RRID:SCR_012761), or present in dbSNP (RRID:SCR_002338) is filtered; however, variants also present in at least twice in COSMIC (RRID:SCR_002260) are rescued for manual review. Variants that appear two or more times in a panel of 150 normal samples sequenced in-house and are not present in COSMIC are also filtered. For copy number alteration (CNA) analysis, the workflow employs an in-house algorithm (RobustCNV) to cover exonic regions of targeted genes as well as select intronic regions. The depth of coverage of these regions is determined by counting the number of reads aligning within defined genomic intervals and then normalized against a panel of normals and corrected for GC bias. The copy number for a segmented genomic interval is calculated as a log2 ratio of the depth of coverage of this sample compared to a panel of normal (non-cancer) samples that are run on the same plate. In parallel (samples not included), additional cases were previously sequenced via the MSK-IMPACT platform. To ensure maximum overlap between the platforms, only single nucleotide variants (SNVs), dinucleotide/oligonucleotide variants (DNVs/ONVs), insertions/deletions (indels), high copy gains (amplifications), and homozygous deletions from 182 genes shared between all versions of both platforms were evaluated in our analyses involving samples from both platforms. To filter pathogenic and passenger mutations, missense, truncating (nonsense, frameshift, and splice site), and in-frame indel mutations from the 182 genes overlapping between the DFCI OncoPanel and MSK-IMPACT panels were subjected to analysis in the Cancer Genome Interpreter (CGI) suite (RRID:SCR_023752) with cancer type set to “Skin Cutaneous Melanoma (SKCM)”. Known pathogenic mutations, defined as those listed as pathogenic in ClinVar (RRID:SCR_006169), OncoKB (RRID:SCR_014782), and/or CGI databases, were included in downstream analyses. Mutations which had not been previously annotated but were predicted to be pathogenic mutations by CGI were also included. Missense SNV mutations called passengers in CGI were subjected to another round of analysis in the ChasmPlus suite with cancer type set to SKCM. Missense mutations called pathogenic via this approach (p < 0.05; FDR Q < 0.3) were salvaged and included in further analyses. All other mutations were excluded. TERT promoter mutations were evaluated via FATHMM-MKL. Mutations called pathogenic via this tool were included in downstream analyses. Clinical data for these cases is available in a supplementary file from the associated publication. OF NOTE, THE RAW SEQUENCING DATA FROM THESE SAMPLES CANNOT BE MADE PUBLICLY AVAILABLE BECAUSE THE RESEARCH PARTICIPANT CONSENT DOES NOT INCLUDE AUTHORIZATION TO SHARE IDENTIFIABLE DATA.