Best practice for variant calling on Ion Torrent data with GATK

zhangx

Hi,

I wonder how well GATK works with Ion Torrent data. Is there any recommended practice to handle Ion Torrent data, especially SNP and indel calling, with GATK?

Thanks, XZ

Mark_DePristo

This is a very good question. We do have reasonably extensive experience with GATK on Ion Torrent (PGM and Proton) and we've been meaning to put together a best practices doc for Ion Torrent. My two cents: we can do reasonably well for SNP but indels are I think impossible today given the homopolymer error mode of that sequencers. I have yet to see any compelling evidence that indels can be done effectively using any approach -- not just GATK -- including the example results on the Ion Community itself.

zhangx

Thanks for your prompt response, Mark. I understand that the variant caller developed by Ion Torrent uses a modified version of GATK, which includes a flow space modelling function to tackle the homopolymer issue. Do you have any experience with that? If I want to call INDELs with on Ion data with GATK now, would you recommend using the current INDEL pipeline -- indel realignment, UnifiedGenytyper with glm set to DINDEL and hard filtering?

Mark_DePristo

All good questions -- things we'll address on the best practices entry.

Niels

Is there more known about using GATK for variant calling SNPS and Indels?

Geraldine_VdAuwera

Niels, if you mean to ask about calling variants on Ion Torrent data, we have not yet had the time to produce that documentation.