Manual Polygenic Risk Score calculation

Question

[Roy]

Hi all,

I am attempted to calculate PRS manually, and I'm very close to to obtaining a score.

To recap what has been done, I have a patients individual in which I annotated their VCF with RSIDs. From there, I went to PGS catalog and found the respective disease (ex. breast cancer) and obtained a paper's scoring file (screenshot attached below).

This scoring file is missing an RSID column, but essentially I would see if this scoring file's RSID matched the patient's RSID, and if so, I multiply the number of copies for the effect allele (black circle) to the red circle, and do the summation of all alleles that this patient had that matched the scoring file. In the end, the number obtained will be compared to the PRS value found in the paper, and maybe plot this value in the percentile PRS curve if they give this.

My question resides in the yellow circle, what if my patient matches the RSID but the allele he contains is the "other allele"? Some variants provide protective characteristics, so is that what this allele is signifying (aka a negative number will be obtained for this allele).

Thanks for any input, Roy

[ashaneev07]

Hello,

Did you calculate the score successfully? I'm also facing issues with prs score calculation.

Answer 1

[Sam]

Same allele = weight of 1, different allele = weight of 0. You can read more here: https://choishingwan.github.io/PRS-Tutorial/plink/#generate-prs

Though I don't think PRS is suitable for one sample analyses. As of now, it is still mainly a research tool for population analyses