how to combine gwas and functional betas from one data set

[bnj50]

Hi My assumption of using polyfun, is to get a better PRS prediction when we combined the betas from gwas to betas from functional effects. To do this , in my discovery gwas set, first i applied the basic python extract_snpvar.py, with this simple command:

python extract_snpvar.py --sumstats --out

As a result, the SNPVAR column has been added to my sumstats file as shown below. Now i want to combine the beta from discovery GWAS (BETA column) to SNPVAR to get a combined beta value.

CHR BP SNP A1 A2 SNPS BETA SE p_value SNPVAR 1 752566 rs3094315 G A rs3094315 -8.3089e-03 2.0156e-02 6.8017e-01 6.2623e-08 1 779322 rs4040617 A G rs4040617 -2.8307e-03 2.4648e-02 9.0857e-01 4.0733e-09

After I get the combined beta, then i plan to run PRS-cs method using these new "combined betas" as discovery set.

can you provide advice and help on this (what command to use) and whether my strategy is correct?

[omerwe]

Hi,

I'm not sure how SNPVAR (SNP-specific prior variance) can be used with PRS-CS. You may want to reach out to the authors of PRS-CS to ask them.

You can use a fine-mapping-driven PRS using PolyPred, as described in the Wiki.

[bnj50]

Hi this description is a bit not clear to me. what does it mean fine mapping for each locus and then aggregate them! i have single GWAS data for whole genome. also number 3 option is related to my question (3. Linearly combining the effect sizes of PolyFun and the other method). can we use this approach to combin snp-var value as functional beta to gwas effect size beta? if yes, how, and what is the input format? specifically, the paper mentioned of "poly-pred-P" [=PolyPred-P+, which replace BOLT-LMM with either SBayesR or PRS-CS]

[omerwe]

Hi, everything is described in the wiki and in the paper. If you have a specific question I'm happy to answer, but I can't explain the entire approach of these methods here.