How to include coverage information in vcf

[krawitz]

Hi folks

When we report variants in vcf format a common question is: what is
the coverage profile of the target region? Combining vcf and coverage
information allows to deduce also the genotypes of the target region
without variants based on the reference sequence. This is much more efficient than listing the 0/0 (which is reference genotype) at every non variant position in vcf.

For this purpose one could generate a bed file listing intervals with e.g. >10x and >20x coverage. Alternatively a wig file would be an option.

I was wondering whether there is also a possibility to include this
information in vcf. Maybe in the header? Does anyone have an idea?

thanks!
Peter

[Richard Finney]

Yes you can do it. Our group as done it in haplotype discovery.

The point you are assuming is ...

IF (there's enough coverage for a location) and (the location is not reported in a VCF that is only reporting non-reference calls)
THEN the position was interrogated and you can assume that the call is "reference".

Of course if there is little or no coverage, you will have trouble making a reliable call.

This is do-able, as you suggest, by having an additional wig (or bigwig) file.

There is the tricky situation of ... other samples have heterozygous at a locus but we only have coverage of 4 in the sample we're trying to call with 2 read loci are are non-reference and 2 are reference. Coverage is only four, but evidence is good for "het". But, of course, making a call with coverage 4 and all 'reference" would be bad. (right?). In practice, I've just used a cut-off of coverage 10.

NB: VCF generation programs can sometimes be forced to output calls for every location, not just the non-reference calls.

[swbarnes2]

The DP and DP4 elements in the info column give coverage info at that position. But by its nature, a vcf is only going to tell you what's going on at the variant. Use BEDTools to get coverage stats across multiple regions.