GATK DepthOfCoverage at high depth

[lletourn]

I have a sample on which custom capture was done. Many of the baits have >2000x coverage.

When running DepthOfCoverage initially I never saw values >500x but then noticed I forgot to set the bins bigger. I set them at 20,000 just to be safe, now no values go over 1000...I really don't get why.
I even set a summaryCoverageThreshold at 1200 and it's always 0.

With genome browser I can clearly see some baits at >2000 even some at 5000.

Any ideas

Here's the command:

java -jar /data/solexa/aligners/GenomeAnalysisTK-1.0.5777/GenomeAnalysisTK.jar -T DepthOfCoverage -R human_hg19.fasta -I sample.bam -o sample.targetCoverage -L all.interval_list --minMappingQuality 15 --minBaseQuality 10 --omitDepthOutputAtEachBase --logging_level ERROR --summaryCoverageThreshold 30 --summaryCoverageThreshold 50 --summaryCoverageThreshold 200 --summaryCoverageThreshold 500 --summaryCoverageThreshold 700 --summaryCoverageThreshold 1000 --summaryCoverageThreshold 1200 --summaryCoverageThreshold 2000 --start 1 --stop 20000 --nBins 19999

[lletourn]

I finally found my answer.

By default GATK downsamples by sample. The downsampling coverage is...drum roll...1000.

To fix this set the number of bins *AND* set -dt NONE

Why one would downsample when computing coverage is beyond me.

[spreeth84]

Thanks! looks like this also happens in the Unified Genotyper in reporting the AD and DP numbers. I was struggling to explain why the DP did not match the coverage at the position and if the quality filters were being too stringent!