We have filtered our VCFs and used the extra perl scrips included with somaticsniper highconfidence.pl) to select for those variant calls that we believe are quality somatic mutations and possible drivers of cancer in our model. This obviously eliminated a great deal of false positives, and I'm wondering if this affects calculating the Background Mutation Rate. Obviously we don't want to feed it a huge number of false positives, but if we hone down our VCFs to certain high quality calls, then make a MAF and run this BMR set, sometimes the overall_bmrs file has 1 or 2 mutations for a sample. Would it be prudent to use a less strict filtering step to make a MAF for BMR calculation?

Thank you