Mutation calling and quality control

Based on the best practice procedures for sequencing alignment and quality control⁵⁶, somatic mutations were called by MuTect2 using genomic references from the Broad Institute⁵⁷. We created a Panel of Normals (PoN) by aggregating all the normal samples so that we could remove common germline variants as well as commonly noisy sites (e.g., mapping artifacts or other somewhat random but systematic artifacts of sequencing). This PoN also served as the normal sample for P3 and P5 since they did not have matched normal samples for somatic calling. We applied the default filter to conservatively select somatic calls with confidence.

Final mutation calls were selected through a stringent filtering process and functionally annotated by ANNOVAR⁵⁸.

We applied the following filtering criteria for somatic mutation calling: (1) read depth > 25; (2) mutant AF > 0.05 in tumor samples; (3) corresponding allele frequency <0.01 in matched normal samples (if present); (4) mutations listed in 1000 Genomes Project⁵⁹ or Exome Sequencing Project⁶⁰ removed.

The following filtering criteria were applied for germline variant calling: (1) read depth ≥ 50; (2) genotype quality score ≥ 30; (3) allele fractions ≥0.3 and ≤0.7; (4) multiple-allele variants removed; (5) variant quality score recalibration (VQSR) ≤ 97.00; (6) variants in segmental duplication removed⁶¹.

We validated the quality of our somatic mutation calls using methods that we have previously established⁶¹. Briefly, when running Mutect2 in patients with matched normal samples (P1, P2, P4, and P6), we performed the same pipeline and filtering criteria but switched the normal and tumor samples. The mutation calls that passed the criteria are declared as artifactual mutations. If there were major artifacts in FFPE samples, we would be able to call artifactual mutations in matched normal samples since they were also FFPE samples.