search

ttimbers / SKAT_NGS-2015

Other
4 stars 14 forks source link

Parallel SKAT-O run per gene? #5

Open gpcr opened 8 years ago

gpcr commented 8 years ago

@ttimbers : thanks for the wonderful tutorial on running SKAT. 1. Is there a way to run parallel SKAT-O per gene runs on multicore (32 cores with 256GB RAM) computer? I am planning to run on 1000's of genes on many (hundred to 1000s) of phenotypes? and also how to deal with multiple currection Pvalues in this situation

my current plan would be: "on the fly" recreate .bed, .bim .fam per gene and run separate SKAT-O instance per bundle of phenotypes on each processor core..... but I am looking for more efficient ways?

ttimbers commented 8 years ago

@gpcr sorry for the much delayed response, I was travelling when I received this and missed coming back to it until now.

In all honesty, I am not aware of a better way than you suggest. As for how to deal with the multiple correction for p-values, what I might try after your parallel processing is to collate all your p-values for one phenotype into one file, load then into R (or whatever program you want to use) and do an FDR on them. Repeat this for each phenotype.

Have you thought of doing some form of multi-variate GWAS analysis with the multiple phenotypes instead of doing them individually? In theory, that should give you more power.