I remembr a discussion here recently regarding "p-hacking". With so many thousands of genes, aren't a fair number going to be correlated or predictive with statistical significance by sheer probability?