I'm searching for genes that are correlated with a specific phenotype through GWAS, but I'm currently struggling with the adjustment of the p-value to reduce the amount of false positive.
I'm using multiple phenotypes and for each phenotype I have got p-values for every SNP. Since it is a multivariate analysis, I wanted to merge the p-values of the different phenotypes and decided to do so by Fisher's method. However my phenotypes are correlated so the merged p-values are artificially lower so a simple bonferonni correction for my significance threshold doesn't seem correct. To counteract for this, I wanted to use a permutated dataset, in which I switch the labels.
However for me it is unclear to do so. Is there somebody who could advise me how to do this or send some good papers explaining this?