merging issue between normal and tumor samples VCF for deepsomatic v1.10

[ChristopheLegendre]

A bug in DeepSomatic v1.10 tool itself prevents bcftools merge command to merge vcf files because of a wrong value assignment for the Number.
Deepsomatic sets the Number value to R instead of A for the tag NAF.

Please fix the issue within vcfMerger2 if possible.

Thanks,
GrandCanyon JetstreamProd User

[PedalheadPHX]

Can you explain this issue a bit better or provide the example VCF line, not sure what you mean by Number to R not A, when they are not numbers

[ChristopheLegendre]

The issue is not vcfMerger2 itself. The issue is because of DeepSomatic itself. In order to not wait for a fix by the DeepSomatic team, I wanted to fix the issue at the vcfMerger2 level.
I added a lot of comments in the code regarding this issue. Let me add more info here as well as requested.

In the VCF header, DeepSomatic declares the tag NAF as Number=R, implying that the tag must have values for the REF allele and all the ALT alleles.
Unfortunately, NAF representing the Allele frequency for all the ALT alleles present in normal, must only show the values for the ALT alleles and not the allele frequency for the REF. Therefore, the tag should have been declared as Number=A, not "R".

Here the fix I implemented in vcfMerger2:

##FORMAT=<ID=NAF,Number=R,Type=Float,Description="VAF of ALT alleles in the normal sample.">

fix the line with:

##FORMAT=<ID=NAF,Number=A,Type=Float,Description="VAF of ALT alleles in the normal sample.">

Examples:
AF does not need to give frequency for the REF since it can be calculated from the ALT AF
AD is another example where Number must equals R and not A, because the AD value for the REF cannot be calculated or inferred from the ALT information.

Descriptor	Applies To	Number of Values	Example Field	Example Values
Number=A	Alternate alleles only	= # of ALT alleles	AF (Allele Frequency)	AF=0.333,0.667
Number=R	All alleles (REF + ALT)	= 1 + # of ALT alleles	AD (Allele Depth)	AD=10,30,40

bcftools merge command looks at the Number information in the header line for the TAGs and check if the value reflects the definition. If not, bcftools merge fails. WHen values are missing and a dot exists, bcftools does not care about the Number definition and keep the missing value as is, in our case here with the normal sample, a dot.

[PedalheadPHX]

Is this a known issue in deepsomatic on their issue board? if not please report it

[ChristopheLegendre]

No it is not. But ...
We have reported it as an issue on the DeepSomatic github repository: issue number 91.

Note: I can put the full link to the issue here to get some advertisement for vcfMerger2 because that would associate vcfMerger2 to deepsomatic.

Let me know. If not we will close the issue here.