Hi all,

I am encountering the following problem when developing methods for calculating polygenic risk score. Basically, some SNPs in my GWAS summary stats are reverse complement to those in the reference panel. I need to use reference panel to estimate LD so I need to align alleles first. Suppose a SNP has Effect/Noneffect as A/G and Z-score of 1.0 in summary stats, and Effect/Noneffect as T/C in the reference panel. To align the allele, this post suggests one should change the alleles A/G to T/C and Z-score to -1.0. My question is if a SNP is A on one strand, then it must be T on the other strand, so why do we need to flip the sign of Z-score? In practice, I do find that flipping the sign is necessary for my algorithm to work.