Saturday, May 22, 2010

Performing a pathway analysis of GWAS data

So you have the results of a GWAS analysis and want to see if any specific pathways are enriched. This is a summary of how you could do that using one of a number of tools available, the SNP ratio test (SRT)

This assumes you have the data in PLINK format.

(1) make 1000 randomized alternative phenotypes:
perl mygwas.fam 1000

(2) run e.g. association on the original GWAS dataset, e.g.
plink --bfile mygwas --assoc --out mygwas_res

(3) run association on all the randomized alternative phenotypes, e.g.
plink --bfile mygwas --pheno mygwas.fam.altpheno.1000.txt --assoc

**NOTE: you can run --assoc or instead --logistic with --covar etc. to run regression with covariates, or --mh to run the Cochrane-Mantel-Haenzel test.

(4) pull out SNP and P-value information:

(5) run the SRT on the original dataset:
perl 0.05 mygwas_res.assoc.forSRT

(6) how many SNPs are at the top? (corrects for inflation of p-values):
perl mygwas_res.assoc.forSRT 0.05

(7) run the SRT on the randomized phenotypes (number is the result from (6)):
perl 12075

(8) get the p-values for each pathway:
perl mygwas_res.assoc.forSRT.p0.05.ratios 0.05

See for more information.

Have fun!

No comments:

Post a Comment