How pgs-calc identify multialelic variants?

[agnis-iohana]

Hi!
I want to calculate pgs for a trait with multialelic variant reported by literature, but I don't know how to include this type of SNP in my score file. How pgs-calc identify it in my vcf? Is there a way to standardize this information in my score file?
I tested a multialelic variant like this:
rsID chr_name chr_position effect_allele effect_weight hm_source hm_rsID hm_chr hm_pos hm_inferOtherAllele
rs63751012 3 37035147 A 0.066 ENSEMBL rs34579442 3 37035147 G,C,T

my vcf report this variant like this:
3 37035147 rs63751012 G A,C,T . PASS . GT:GQ

But this SNP appears as a mismatch and the multialelic field is shown as false in the output, however I set --keep_multiallelic parameter as true.

Thanks!

[nebfield]

Thanks for your question and sorry for the late reply. The GitHub notification got lost in my emails 👻

If I understand correctly you want to use a custom scoring file (not published in the PGS Catalog) that contains a variant with multiple other alleles.

We have some documentation about creating custom scoring files. The data structure is different to the example you described:

https://pgsc-calc.readthedocs.io/en/latest/how-to/calculate_custom.html#scorefile-setup

However, in v1.3.2 we don't support variants with multiple other alleles encoded. These variants are removed internally during quality control steps. This is why the variant is labelled as not multi-allelic. Correctly handling multi-allelic variants can be complicated so this might change in a future release.

One potential workaround would be to test making a custom scoring file using only effect_allele data (leaving other_allele empty).

Thanks for your interest in pgsc_calc 🚀