ClusteredReadPosition

Detect clustering of variants near the ends of reads

Category Annotation Modules

VCF Field INFO (variant-level)

Type ActiveRegionBasedAnnotation


Overview

This annotation detects clustering of evidence for a somatic variant near the ends of reads. To turn on the annotation and the accompanying filter (clustered_read_position), add --enable_clustered_read_position_filter flag in the commandline.

Statistical notes

ClusteredReadPosition produces four INFO field annotations. At a given somatic variant site, MEDIAN_LEFT_OFFSET is the median of the number of bases from the left end of the tumor read to the variant. MEDIAN_RIGHT_OFFSET is similar, but counts from the right end of the read. MAD_LEFT_OFFSET and MAD_RIGHT_OFFSET measure the median absolute deviations. The median gives us the offset of a representative read, while the median absolute deviation captures the spread. We filter a variant if MEDIAN_LEFT_OFFSET <= 10 and MAD_LEFT_OFFSET <= 3, or if MEDIAN_RIGHT_OFFSET <= 10 and MAD_RIGHT_OFFSET <= 3.

Caveat

ClusteredReadPosition is available with MuTect2 only

RelatedAnnotation

  • ReadPosRankSum is a similar annotation designed for germline variants.

    Return to top


    See also GATK Documentation Index | Tool Docs Index | Support Forum

    GATK version 3.8-0-ge9d806836 built at 2017/07/29 01:40:22.

  •