Why Do We Need To Set Maximal Read Depth In Samtools

Question

[a.a.kechin]

I analyse reads from Ion Torrent platform with bwa-samtools pipeline. I used them in such way:

bwa mem
samtools view (convert sam to bam)
samtoos sort
samtools mpileup
bcftools view -bvcg - > var.raw.bcf (call variants as I understand correctly)
bcftools view var.raw.bcf | vcfutils.pl varFilter -D100 > var.flt.vcf (convert bcf to vcf)

And at the last stage I can't understand why do we set maximal read depth. Don't we exclude thereby many SNPs? By what SNPs with high read depth are bad?

Than you.

Answer 1

[Jorge Amigo]

this is just a resources-saving parameter. if you have plenty of coverage, selecting reads at random up to a maximum number should end up with a representative bunch of reads, enough to perform the variant calling but being much less hard to handle. another example of calling variants considering not all the reads would be GATK, which includes this parameter active by default, although it increases the threshold up to 250. but if you have plenty of time and memory, or if your experiments are not that big, if you feel like you don't want to leave your results in hands of randomness you can just process all the reads. that's what we do when we're able but, to be honest, we haven't seen any better calls with this procedure. for us it's just a way of having a measurement of the exact coverage at that site just sitting next to all the site's details instead of having to query the bam file again later on.