The traditional approach in Mutect and HaplotypeCaller is to filter reads based on simple properties, such as low mapping quality, lack of mate, etc. This still leaves a lot of fairly bad reads. These probably don't affect calls because they don't support any allele well according to PairHMM. Still, they probably waste a lot of time by generating unnecessary paths in the assembly graph.
Some possible filters include an excessive number of different CIGAR elements, and too many low-quality bases.
The traditional approach in Mutect and HaplotypeCaller is to filter reads based on simple properties, such as low mapping quality, lack of mate, etc. This still leaves a lot of fairly bad reads. These probably don't affect calls because they don't support any allele well according to
PairHMM
. Still, they probably waste a lot of time by generating unnecessary paths in the assembly graph.Some possible filters include an excessive number of different CIGAR elements, and too many low-quality bases.