Calculating capture efficiency

[T199Y]

Hi all,

I want to calculate what percentage of my sequenced custom capture is actually off-target and I can't get my head around how to do this - does anyone have any tips?

Many thanks.

[Heisman]

A couple of ways you can do this.

1. The easiest would be to create a reference sequence for your custom capture, and then align your data to the whole genome and your custom reference sequence (give your custom reference sequence 500 bp flanking region). Then see how many reads align to the whole genome but don't align to your custom capture. There may be a bit of ambiguity due to some reads aligning to the custom capture that don't align to the whole genome if they have high homology to multiple portions of the human genome, but this won't be a big issue from my experience.

2. If you have aligned to the whole genome and don't want to go through the trouble aligning to the custom reference sequence, you can run GATK depth of coverage for the whole genome and also for the specific targeted intervals (again, + and - 500bp flanking regions). Then compare the total amount of coverage.

[Jon_Keats]

Try Picard (http://picard.sourceforge.net/comman...ulateHsMetrics)

[swbarnes2]

I use BEDTools. Get a .bed file of your capture regions. Align to the whole genome (which is in general a better idea than just aligning to target region), then intersect your whole genome .bam with the .bed file. You might pad the bed file first, though if you use the exact coordiantes, you will not lose reads that hang off your target region.

Then count (with samtools flagstat, or something like that) how many lines are in your bam before and after filtering.