Improving the prediction of the functional impact of cancer mutations by baseline tolerance transformation

Table 2 Composition of the datasets used as proxies to compare the performance of transformed and original scores at assessing the functional impact of cancer somatic mutations

Name	Source	Positives	Negatives	N positives	N negatives
COSMIC2+/1	COSMIC	Mutations that appear in 2 or more samples	Mutations that appear in 1 sample	4,012	39,854
COSMIC5+/1	COSMIC	Mutations that appear in 5 or more samples	Mutations that appear in 1 sample	1,480	39,854
COSMIC2+/Pol	COSMIC/ HumVar [2]	Mutations that appear in 2 or more samples	Known polymorphisms	4,012	8,257
COSMIC5+/Pol	COSMIC/ HumVar	Mutations that appear in 5 or more samples	Known polymorphisms	1,480	8,257
COSMICD/O	COSMIC	COSMIC mutations included in the manually curated list of drivers used to train CHASM [5]	COSMIC mutations without the positive subset	2,185	41,681
COSMICD/Pol	COSMIC/ HumVar	Mutations included in the manually curated list of drivers used to train CHASM	Known polymorphisms	2,185	8,257
COSMICCGC/ nonCGC	COSMIC	COSMIC mutations in genes included in the Cancer Gene Census [13]	Non-recurrent COSMIC mutations in genes not included in the Cancer Gene Census	4,685	35,907
WG2+/1	Pooled cancer somatic mutations	Mutations that appear in 2 or more samples	Mutations that appear in 1 sample	1,031	26,025
WGCGC/ nonCGC	Pooled cancer somatic mutations	Mutations in genes included in the Cancer Gene Census [13]	Non-recurrent mutations in genes not included in the Cancer Gene Census	1,412	24,837

HumVar is a dataset of disease-related SNVs and neutral polymorphisms [2]. WG (whole genome) is a dataset of somatic mutations pooled from different tumor exome-sequencing projects (Table 1).