Table 1. Multi-step filter selects cancer drivers in genomics datasets with high mutational burden.

Step	Logic	Program	Input	Output	Reference files
i	Cohort selection	TCGA Portal, CGHub, Gene Torrent	tcga_patient_list.txt	wes.bam
ii	Mutation call	MuTect 1.1.428	wes.bam	coverage.wig, call_stats.txt	HG19, COSMIC_54.vcf, dbsnp_132.vcf
iii	Recurrence, evolutionary conservation	MutSig 2.029	patient.maf, coverage.wig	covariates.txt	HG19/.maf
iv	Correction for background mutation rate	InVEx 1.0.19	coverage.wig, covariates.txt	significant_mutation_burden.txt, qq.png	HG19/.maf, PPH2, nucleotide_classes_HG19.txt, COSMIC, genePeptideFile_HG19
v	Mutation Signature, UV induced damage	Text editor	patient.maf	sorted_transitions.maf	nucleotide_classes_HG19.txt, uv_transitions.txt10
vi	Structure-activity-relationship	SWISS-MODEL 8.05, TMpred, 25.0	structure.pdb	model.pdb, tm_model.txt
vii	Pathway enrichment, Mutual exclusivity	GSEA 2.1.012, MEMo 1.111	tcga_patient_list.txt, scna.txt, amp_del_gene.txt, covariates.txt, coverage.wig	modules.txt
viii	Recurrence within PAN-Cancer TCGA	TCGA	covariate_target.txt, patient.maf	covariate.maf

Somatic mutations of driver genes are called after i) cohort selection, ii) mapping of human genome and patient specific somatic references, iii) assessment of recurrence, evolutionary conversation, and iv) basal mutation rate based on frequency of mutations of introns vs exons. This first set of filters i)-iv) is necessary and sufficient to identify statistically significant enriched somatic mutations of driver genes in any dataset with high mutational burden. In a genome-wide sequencing experiment with a goal to find cancer drivers, an additional level of filters v)-viii) is advantageous. Relevance of mutations is assessed by v) nucleotide signature, vi) structure activity relationship, vii) pathway enrichment and mutual exclusivity to known cancer drivers, as well as viii) recurrence in other cancer tissues.