With Dr. Teresa Head-Gordon at Lawrence Berkeley Labs, I hacked up (in C) a feed-forward, back-propagation artificial neural network for the purpose of protein secondary structure prediction. We saw some modest improvement by "hard-wiring" expermintally observed sequence-structure relationships into network components. Extensions to this work, including tertiary structure predictions based on folding of predicted secondary structure elements, have also been examined in the Head-Gordon lab, along with other cool stuff.
Last updated 7/31/2000.