pileup mapping quality cap and SNPs?

[MatthewVC]

I am performing some analyses for a sequencing project and am generating pileups from .bam files in Galaxy. In my pileup files afterwards, I’ve found a few places where the sequence for a position includes two “Y”s flanking a T. In this position, the coverage drops very low (to roughly 30 from ~8000).

When generating the pileup, I set the cap for mapping quality to 60. I am thinking this may have caused the reads mapped in this location to be below the threshold due to the two SNPs in close proximity making the mapped reads look “poor”. I wanted to see if my logic was correct here. If I were to drop the mapping quality cap, what is generally considered a fair value to use? As I understand it, a mapping quality of 60 means that there is a probability of 0.000001 that the read is mapped wrong. If I were to set it to 30, with a probability of 0.001 that the read is mapped incorrectly, would that be considered not stringent enough in an analysis such as this? I would think that with such high coverage that this would be acceptable, but I’d like the advice of someone more versed in these types of analyses.

Thanks,

Matt

[MatthewVC]

I redid the pileup with a lower mapping quality cap (40) and that did not change the read or coverage values for these locations. I also tried changing the base quality cutoff when filtering the pileup, lowering it as far as 10, and while that increased the coverage at those locations, it was still very low (~40ish).

I wanted to show some pictures of my problem, to see if that might help someone point me in the right direction.

I opened the .bam file with IGV to look at the alignment and in IGV, the coverage at that location is listed as ~39,000, with 99.99% of those reads being "C" (picture A). When I look at the filtered pileup however, the coverage is listed as 32, with 21 "C" reads and 11 "T" reads (picture B). I'm really puzzled as to why the pileup I'm generating and filtering is showing such low coverage at this particular locus. I'm seeing similar problems at this locus and others across 4 different samples. Any ideas?

[MatthewVC]

Anyone have any insight into what might be causing this problem?

Thanks