CLC BIO for detecting indels in ION Torrent PGM data

[ddaneels]

Hello everyone,

I was wondering if someone has some experience with using the CLC BIO genomics workbench to call indels in ion torrent PGM data? Did you encounter a lot of false positives in homopolymer regions?

I heard that it might be a good idea to remap your data with different settings (different penalty values for gap extension, gap opening, ...) before you search for indels? Any thougths on that?

[jonathanjacobs]

We use CLCbio - and the current version (both 5.5 Beta and 5.1 Stable) has options for homopolymer filtering for SNP/Indel calling with 454 and Ion Torrent data. We haven't verified these indels by Sanger (yet) - but if you disable the homopolymer filter you will see a lot more false positives pop up in your results. The filter CLC has developed for homopolyers works pretty well in my experience.

Try doing your read mapping with tmap3 on the Torrent Server - export SAM/BAM - and do the variant/indel calling using CLCbio and compare to read mapping done with CLCbio's mapper. The results are generally pretty consistent.

Remapping with different settings I don't think is a good idea. I'm assuming you are suggesting making gap penalties for opening/extension higher? This is going to muck up the read mapping and you'll likely get lower coverage overall. If I were to do it, I would want it to be done empiraccly: i.e. vary each penalty independently and measure the effects across a couple of targeted regions that both include/exclude homopolymer stretches in the reference genome. This way you could see what the effects are of playing with your gap penalties a bit more objectively.

What settings are you using for read mapping? What's your desired coverage? What is the genome you are mapping to? Also - how big are these InDels?

The only way to know for sure is to follow up with Sanger sequencing or another NGS approach. This is largely true for any NGS platform.