Read Depth in vcf (samtools / bcftools)

[Marie_Noir]

I generated one vcf file from 4 different bam-files (4 samples) and had a look at some variants in the vcf and in the IGViewer (IGV).

I don't understand why the value for DP in the formatstring (i.e. for one sample) often differs (is less) from what I can see in the IGV. Meaning, vcf tells me a certain variant in a certain sample has read depth (DP) of 3 but I see more reads covering that position in the IGV. I thought that maybe bases of bad quality where left out, but reads and base qualities are good. Is there any other measure that I don't know yet which filters out certain reads / bases from being reported in the DP field in the format string (and consequently is not used for genotype assignment)?

Any hint is appreciated.

[baika]

I also had same question when I started my analysis. Then I figured out that DP (Depth of coverage) is not exactly the "counts of reads" at that position. Please see following links and I am sure you will get the answer:

http://www.illumina.com/Documents/products/technotes/technote_coverage_calculation.pdf

http://www.illumina.com/truseq/quality_101/coverage/coverage_distribution.ilmn

Hope it helps