P-value histogram interpretation

Question

[Inbalor]

How I should interpret a histogram of P-values (coming from multiple test) that is increasing in the left side?

Thanks

[Alex Reynolds]

http://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002106#pbio-1002106-g002

More generally:

http://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002106

Answer 1

[Devon Ryan]

That suggests that you have some actual significant findings. Under the null, we expect a roughly even distribution of p-values.

[Inbalor]

Thanks to reply my question.

I understand that if the proportion of genes with p-values < 0.05 is higher that what you expected is because you have significant results. However, what happened if you have less proportion of genes with p-values < 0.05 that what you expected at random. Why it indicates that your results are significant. Sorry, could you explain me better. Thanks again!

[Devon Ryan]

Perhaps you're plotting things backward from the rest of us, though. If you're seeing significantly fewer significant p-values than expected even under the null then there's often a confounder that's been omitted from your design (at least in my experience).