Post by francesc on Jun 16, 2021 15:16:15 GMT
I am trying to test the effect of changing the effect size written in the GCTA causal variants file (--simu-causal-loci) when simulating binary phenotypes, while maintaining heritability (--simu-hsq) and disease prevalence (--simu-k) constant.
By effect I mean the number of individuals labelled as cases and controls in the output phenotype file, that carry or not the causal alleles. I am expecting that no effect (effect size = 0, odd ratio = 1) would result in an equal ratio of mutated/wild-type individuals in cases and controls; whereas effect sizes significantly bigger than 0 (odd ratio > 1) will result in a higher ratio of mutated/wild-type individuals among cases than controls.
However, the effect size specified in the causal variants file (--simu-causal-loci) had no effect on the distribution of mutated individuals and wild-type individuals labelled as cases and controls by GCTA in the phenotype file. The value that had the biggest and proportional effect was heritability.
As a starting point, I included a single causal SNP in the causal variants file (--simu-causal-loci).
I was wondering if there is a way of testing varying effect sizes when simulating binary phenotypes with GCTA.
This is the command line I used:
gcta64 --bfile /data/mtb_clade.rf --simu-cc 300 300 --simu-causal-loci /data/mtb_clade.isoniazid_causal_mutation.txt --simu-hsq 0.5 --simu-k 0.5 --simu-rep 1 --out /data/mtb_clade.gcta
I am attaching the results as an image.