Negative Polygenic Risk Scores

[Mo Abdulkadir]

I ran PRSice and found that nearly all if not all individuals have a negative polygenic risk score. I was wondering if this is correct or that I am overlooking possible data cleaning issues? If these negative scores are correct how do I interpret them?

I ran the following command:

R --file=PRSice_v1.25.R -q --args \

base base_dataset \

target target_dataset \

covary F \

clump.snps F \

slower 0 \

sinc 0.1 \

supper 0.5 \

plink /sara/sw/plink-1.90b3y/plink \

cleanup F \

figname run_1

[Jonathan Coleman]

Negative polygenic scores are potentially valid. The score is the average of all the weighted test allele counts (with missing genotypes imputed as the mean frequency in the target dataset). If the log(ORs) from the base dataset are predominantly negative (i.e. most test alleles are more common in controls than in cases), then the score will tend to be weighted negatively. This isn't really an issue - in a within-trait analysis, we should still expect cases to have higher scores than controls, because they carry fewer protective alleles.

Check PLINK's score function (on which PRSice is based) to understand further: https://www.cog-genomics.org/plink/1.9/score 

[Mo Abdulkadir]

Thank you Jonathan. It makes more sense to me now.