search

privefl / bigsnpr

R package for the analysis of massive SNP arrays.
https://privefl.github.io/bigsnpr/
196 stars 44 forks source link

Question about LDpred2 on PRS #484

Closed jueqiw closed 8 months ago

jueqiw commented 8 months ago

Hi Florian,

Thanks a lot for this great work!

I am following these two tutorials to compute the PRS using LDpred2: PRS-Tutorial and LDpred2.

However, I got this histogram of the two binary classes using the infinitesimal method on LDpred2: image

The increase Nagelkerke $R^2$ = 0.6741554 - 0.001468591 = 0.6726 looks good, but you can see from the histogram plot that the PRS is from -32 to -26. I wonder whether this is normal or not for the LDpred2 result. This is the code I use. Could you let me know whether I made any mistake or not for the result?

Thanks for great work again!

Jueqi

privefl commented 8 months ago

No, I would not worry about this range. If you need something very well calibrated, I think it should be recalibrated using another dataset.

PS: please only use the official, up-to-date tutorial.

jueqiw commented 8 months ago

Thanks for answering my question! I will check the newest official tutorial again to make sure everything is correct!