Q-Q plots with GWAS p-values using R (needs confidence band)

Question

[joshf]

I'm trying to create Q-Q plots with GWAS p-values using R but it hasn't really been working out. I need for them to be on a -log10 scale, and for them to have a confidence band/interval.

I tried out:

• qqman - works great and it's simple but it does not provide a confidence interval function> code: qq(*P-values*)
• snpStats - has the confidence interval but the X and Y axis are chi-squared values. I'm not sure if there's a way to get -log10 P-values? The plots look great though. code: qq.chisq(-2 * log(*P-values*), df = 2, pvals= TRUE, conc=c(0.05, 0.95))​
• car - the plot looks a little....odd. Maybe I'm just using it wrong but I don't like the distribution of the dots. Why does the X-axis start at -4 instead of 0? code: qqPlot(-log10(*P-values*), envelope=0.95)

I'm hoping the colorful examples would help me get a detailed answer lol

Answer 1

[Kevin Blighe]

Old thread but some tips for others arriving here:

There is indeed something wrong with your distribution. Possible problems:

• your dataset is unbalanced
• your dataset is under-powered
• the statistical test that you employed is not suited to the data distribution that you have
• any combination of the above

Kevin

PS - the bottom plot has X-axis starting at -4 because it is plotting quantiles.