I remember when Len Adelman came to my undergrad institution in '94 or '95 and gave a seminar on his DNA computer. We (biologists) thought it was a neat trick, but not particularly interesting- it was always obvious to molecular biologists that a sufficiently complex set of protein operations could be used to store information and compute over it.<p>What Adelman didn't understand was how poorly his system scaled. To solve problems in his system, every molecule encoding data had to encounter and properly select its mate- that's fine when you have a small number of items in a test-tube volume. But combinatorial growth meant, that to encode problems that are interesting, you would have to use bathtub-sized reactors, and the solution would take a long time, and errors would be hard to manage.<p>People got slightly better but these systems still aren't really used for anything. I'm always hopeful somebody will use a system like this to discover something new and interesting, but most biologists just aren't open to using them (it's a risk, if you tried to publish a major finding using this technique, your reviewers would probably reject the paper).<p>Like material science, material biology has a ton of potential but the followup after discovery tends to be really limited.