I recently read a nontechnical article describing a neural network that had
been trained to predict tertiary protein structure from the primary sequence.
As I recall, there was little information about the extent to which this
network was able to make correct predictions given primary sequences it had
never "seen" before. So I wonder: Has anyone followed up on this research?
Has anyone replicated the result? Does the method show any promise for
predicting tertiary structure correctly, given an arbitrary primary
sequence? If so, it would seem to be an important result, since neural
networks run far more quickly than molecular dynamics simulations of