Indel vcf Processing

Question

[hdtms]

I'm processing Indels from a vcf file, using vcftools I will consider a Min mean DP of 5 but I do not know which Max mean DP to consider. I also do not know the window of Indel length that is usually considered

Answer 1

[Kevin Blighe]

There are no set standards for what you seek.

If you already have a VCF, then the most important part of identifying indels, i.e., variant calling, would already have been performed. Check the parameters that were used during the variant calling process, and also which variant caller was used. In many situations, the variant caller will only make calls from 100 reads over each base (even if there may be 2,000 reads).

5 is very low to use as a cut-off, particularly for indels. Again depending on the variant calling process, going by 5 will result in many false-positive calls. For indels, I would aim for at least 30 read-depth. Do not consider a maximum read-depth from the VCF, as, as I have mentioned above, the variant caller will most likely have dealt with large read depths.

The maximum size at which you can call indels from 150bp paired-end illumina reads is ~22bp. If you have 75bp single-end reads (like with ion torrent), you will struggle to call indels larger than this size.