UnifiedGenotyper between tumor and normal

[kasthuri]

I have 8 tumor and normal samples (16 altogether). If I want to compare the SNP's between them (8 Tumor vs 8 Normal), should I run the UnifiedGenotyper separately twice (one run for Tumor samples and another for normal) or can I run them altogether once with 16 sample bam's. What could be the best way to analyze them?

Thanks.

[fpepin]

Are these matched normals or you have 16 individuals in all?

Running in all together would get you an average of everything, which would be a bad idea. Unless you think that your SNP of interest would be in all tumors, even putting these together would be inadvisable.

GATK specifically says to be REALLY careful about just running the genotyper and removing the SNPs in the normals. It might be ok if you're looking for a few interesting mutations that you can verify manually, but it's not ideal for large-scale efforts.

The authors of GATK say that they have their own effort underway, but it's not public yet. I know of another yet-unreleased effort but I don't know of of any publicly available genotyper for tumor-normal pairs.

[kasthuri]

Thank you fpepin. These are matched normals of 8 pairs. I guess I will run them individually as it is advised.

[pravee1216]

Running Unified Genotyper on individual samples and identifying true somatic calls would be difficult as you may require a statistical score/probability value and variant_allele_frequency (VAF) to support the calls. I see MuTect is one you may try but the tool is still not available to public. You may also look at VarScan, SomaticSniper, Joint-SNV-mix.