A bioinformatics lab I worked at developed a novel method of cancer drug prediction based on comparative genetic network expression from microarray data of normal and cancerous cell lines. First predictions were promising and small scale trials were done with the novel compounds. Very impressive results and a good paper resulted from this which supported further grants for larger trials. Then the lead researcher died unexpectedly.<p>Source code for the original prediction algorithm? Nowhere to be found. Intermediary data? Nada. We spent weeks combing through desktop, personal, and distributed computers trying to find that code to no avail. No more cancer drug experiments. The worst part to me was that this method looked for pre-existing naturally occurring compounds that were merely lacking in cancerous cells - hence reintroducing them would have much lower theoretical toxicity than other anti-cancer compounds. It's still theoretical because proper source control wasn't used.