Secondary Structure Prediction

There have been many attempts over the years to predict protein secondary structure based purely on sequence. The best accuracy reported in the literature to date to my knowledge is ~90%; it has been argued on theoretical grounds that accuracy much greater than 90% is probably not possible due to some uncertainty in precisely how secondary structure is defined. Most such efforts rely on either convolutional neural networks or multilayer LSTMs. This network is a work in progress (hence, the code is not yet extensively commented so may be hard to read). It's aimed at predicting secondary structure using a somewhat more unusual approach. Updates (and comments) coming soon!