Question about ldpred-inf inside XPASS function

[bbitarello]

Hello, I have a #question,

in the output of the XPASS function, your tutorial says mu1 and mu2 are simply the ldpred-inf posterior probabilities for beta. I find that they are different from what I get with the exact same dataset, and hence so are the phenotypic predictions, when I run ldpred1. Am I correct in interpreting that this is because the XPASS function takes posterior probabilities per block whereas ldpred-inf uses a radius (typically between 100-500 SNPs)? Thank you

[mxcai]

Hi there,

Yes, the only difference in our implementation and LDpred1 is that we partition the genome by approximately independent LD blocks while LDpred1 simply uses LD windows defined by a given radius. In practice, we found the predictive R2 of the two approaches are very similar.

Best,
Mingxuan