[stopcodon]

That's why ab initio protein structure prediction is such a sought-after tool in bioinformatics. You can have an incredibly high degree of nucleotide/amino acid substitution between two proteins with the same structure, function and evolutionary history, but current methods of detecting these relationships still largely rely on sequence comparison alone.

[alexholehouse]

I saw Rhiju Das speak last week (really fantastic talk). He's working on some very cool deterministic ab initio folding approaches. http://www.stanford.edu/~rhiju/research.html

[chriscoyfish]

The last time I read into this subject, there were neural network models that achieved around 80% accuracy. There's a research group at my uni dedicated to this research area.

[cing]

I think you're referring to the prediction of "secondary structure" of proteins, which was attempted using neural network models as early as 1989. [1] Predicting secondary structure is child's play compared to predicting the folded tertiary structure of a protein, hence the continued use of insanely computationally expensive ab initio folding efforts like Folding@Home.

[1] http://www.ncbi.nlm.nih.gov/pmc/articles/PMC286422/