PLOS Computational Biology just published a new addition to their popular “ten simple rules” series:
Sandve GK, Nekrutenko A, Taylor J, Hovig E (2013) Ten Simple Rules for Reproducible Computational Research. PLoS Comput Biol 9(10): e1003285. doi:10.1371/journal.pcbi.1003285
This article is relevant to anybody who wants to do computational research. I’ll make it required reading in my lab. For every single one of these rules, I can think of projects I’ve been involved with  that ran into trouble or failed because they violated that rule.
While all of the rules are important, I’m particularly partial to these four: avoid manual data manipulation, record all intermediate results, always store raw data behind plots, and provide public access to scripts and results. They will prevent a lot of headaches for both you and the people coming after you who’d like to build on your results.
 All of these projects were run by friends of friends, of course. None of this would ever happen in my lab. :-)