Hi everyone!
We have just received a new large dataset (ALSPAC dataset) which has been thoroughly QC and imputed against the HRC reference panel. The dataset is UK-based. I've run a quantitative GWAS for two continuous phenotypes and a case-control GWAS all without covariates on SNPTESTv2.5.4. When I got the results the qqplots look very wrong and skewed. I've calculated the lambda inflation factor with the estlambda(gwas$P) from the GernABEL package. The lambda is very very high (3.65 and 3.62 for the two continuous phenotypes and 4.12 for the binary phenotype).
QQ plots:
I have not yet adjusted for population stratification, but since the individuals are all UK I don't think it is the problem, especially since lambda is so high! Do you have any idea of the significance of this behaviour or how to fix it?
I've tried to adjust the P values to lambda, I've looked online and found a solution on R (below), but the qq plot with the adjusted P value is not good either!
chisq <- qchisq(gwas1$P,1,lower.tail=FALSE)
lambda <- median(chisq)/qchisq(0.5,1)
newchisq <- chisq/lambda
gwas1$P_new <- pchisq(newchisq, df=1,lower.tail=FALSE)
Thanks a lot for the help! Filippo
