gatk-3.8/protected/java/test/org/broadinstitute/sting/gatk/walkers
Mark DePristo bf8421eeb7 Fixes GSA-671 / AFCalcResult.log10pNonRefByAllele should really be log10pRefByAllele
-- The current implementation of AFCalcResult contains a map from allele -> log10pNonRef. The only use of this field is to support the isPolymorphic function per allele. The call to this function looks like isPolymorphic(allele, QUAL). The QUAL is a phred-scaled threshold where you want to include alleles where the log10pNonRef >= QUAL (appropriately transformed). The problem is that when log10pNonRef is large, it quickly gets set to 0, while it's complementary log10pRef value has a meaningful log10 value. For example, if log10pRef = -100 (not an uncommonly large value), log10pNonRef = 0.0.
-- In order to preserve precision and allow us to more finally differentiate high QUAL from low QUAL (but still poly) sites we should store log10pRef values instead, and test that log10pRef <= threshold.
-- See https://jira.broadinstitute.org/browse/GSA-671 for more information.
2012-12-07 16:03:40 -05:00
..
bqsr Updating tests to handle bad validation data files (that used the wrong qual score encoding); overrides push from stable. 2012-12-03 22:01:07 -05:00
compression/reducereads Co-Reduction implementation in ReduceReads 2012-11-14 10:33:21 -05:00
genotyper Fixes GSA-671 / AFCalcResult.log10pNonRefByAllele should really be log10pRefByAllele 2012-12-07 16:03:40 -05:00
haplotypecaller Updating the HC integration test md5s. According to the NA12878 knowledge base this commit cuts down the FP rate by more than 50 percent with no loss in sensitivity. 2012-12-03 15:38:59 -05:00