Skip to content

inconsistent allele depth and total depth source #1022

Description

@anoronh4

When combining Strelka and Mutect calls in SomaticCombineChannel, the way annotations are combined is inconsistent. For calls that are shared by both callers, DP comes from Strelka and AD comes from Mutect2. As a result, n_depth is usually different from the sum of n_ref_count and n_alt_count, even in the absence of a third allele. The same thing happens to t_depth, t_ref_count and t_alt_count, although t_depth is overwritten in the downstream process SomaticFacetsAnnotation with annotate-maf-wrapper.R from facets-suite.

The following snippet is supposed to annotate shared calls with Strelka annotations that Mutect does have, due to the pattern +FORMAT:

bcftools annotate \
--annotations ${isecDir}/0003.annot.vcf.gz \
--columns +INFO,+FORMAT,Strelka2FILTER \
--output-type z \
--output ${isecDir}/0002.annot.vcf.gz \
${isecDir}/0002.tmp.vcf.gz

The plus sign means that the TAG is added if the source value is not missing and TAG is not present in the target file. Documentation can be found here. However, bcftools 1.9 in the current container doesn't actually behave this way. bcftools 1.11 does work the way we expect, and presumably newer versions do as well.

As for the over-writing behavior of facets-suite, either both values need to be overwritten, or neither of them. We do not want to change how facets-suite works, but the calculated depth does not necessarily have to be written into the maf.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions