Pan-cancer analysis reveals technical artifacts in TCGA germline variant calls

BMC Genomics

Table 3 Metrics of filter stringency and efficacy

Filter	LOF indel sites	Median LOF indel burden	Fraction discordant indels removed	Fraction concordant indels removed	Indel overlap with ExAC
VQSR 90	6212	53	0.8667	0.4514	0.7079
VQSR 95	9177	59	0.8064	0.3760	0.6776
Hardfilter	24212	91	0.3600	0.0210	0.3527
VQSR 99	26134	98	0.2763	0.1100	0.5394

GATK VQSR 90 is the only filter capable of eliminating the significant association between WGA and LOF indel burden, however; it does so at the cost of over 75% of all LOF indel sites (Additional file 1: Table S10). From this we can conclude that WGA artifactual indels closely resemble true indels, preventing VQSR from selectively removing artifactual indels

ISSN: 1471-2164