winners in SMC (Somatic Mutation Challenge)

Question

[Bogdan]

Dear all,

we are looking forward making a decision about the algorithms to use for SNV detection : please could advise on the top 4-5 winners in SMC Mutation Challenge for SNP/indel detection.

many thanks,

bogdan

[Michael]

Hi tanasa, could you please make your post into a real question by editing it slightly? Do you want us to comment on some of those algorithms, or should we vote? What kind of advice do you require, and what is your data? This will help us to answer your question. Thanx

[Bogdan]

Hi Michael, thanks for the quick reply. Yes, I would need indeed to be more precise with my question. It would be very helpful if you all could please comment on the top performers from SMC (Mutect, Strelka, others) based on your experience: were the predicted mutations by the top performers indeed very well validated in your samples ?. Or if you have any other suggestions regarding other complementary algorithms to use ... Many thanks !

Answer 1

[Chris Miller]

It really depends on the details of your study and the balance of specificity and sensitivity that you're looking for. No caller will get every mutation, and that's especially true at low VAFs.

See our paper "Optimizing Cancer Genome Sequencing and Analysis" for some discussion and some comparisons of variant callers on this an extremely deeply sequenced and very highly validated cancer sample. This figure from the supplement is probably the most relevant, but there are other sections that will be of interest. Note that everything is pretty crappy below 10% VAF, even with very deep (300-1500x) sequencing data. Also look at figure 4 and note that even doing simple intersections of variant caller results can dramatically improve performance.

[ivivek_ngs]

Ah finally a great resource to exploit. Thanks for the paper, I really should not have missed out this paper. Great resource, thanks a lot for mentioning this. Will thoroughly study it and come up with queries if I have.