Nature News & Comment has just published a good comment:
How scientists fool themselves—and how they can stop
The comment goes through a number of the standard ways people fool themselves, but skirts around the most important one in modern biology: failure to correct for testing multiple hypotheses. They mention “p-hacking” as a problem, but their prescription is just “don’t do it” rather than explaining how one corrects for testing many hypotheses.
I think that the comment could have been much stronger if they had gotten some statisticians to provide the real corrective measures needed, rather than just moralizing about how people fool themselves.