bcftools norm convert non-ATGCN IUPAC to N however mpileup and call does not.

[gokalpcelik]

Hi

I noticed that mpileup and call incorporates non-ATGCN IUPAC bases into VCF however bcftools norm expects only ATGCN. Is this an expected outcome? I thought non-ATGCN bases are not allowed in VCF spec? Latest spec dictates conversion of R to A for example though 4.2 does not mention it.

This happens when GIAB reference or any of the GATK resource bundle references are used for mapping and calling.

[pd3]

What version of bcftools are you using? I believe recent versions of bcftools mpileup convert non-ACGT characters to N

[gokalpcelik]

I am on 1.21. I can post a few samples here but I forgot which sample had those issues. I will post as soon as I can track one.

[pd3]

Please do provide a test. I tested myself and it is working as expected 0bce7ab

[sirus20x6]

Related fix in #2527 — the csq codon translation macros now clamp the index for non-ACGT bases (returning 'X' instead of reading past the gencode buffer).