Is SAM-T06 suitable for prediction of transmembrane domains?
SAM-T99 and SAM-T02 have not been optimized for transmembrane predictions. They are “ok” on transmembrane predictions, but not nearly as good as tools optimized for that task. We’ve been told that the TMHMM server is currently the best predictor for transmembrane helices, but we’ve not done any tests ourselves. • SAM-T06 returns what appears to be probabilities a residue belongs to one of seven DSSP defined secondary structure states (the scores sum to 1). Are these numbers really probabilities that the given residue truly is the given secondary structure type (as verified by testing on a validation set), or are the numbers something more akin to probabilities that sequence homologs will have that secondary structure type at that particular location? The probabilities returned by the SAM-T06 server are from neural nets. The neural nets were trained to maximize sum_examples log Phat(correct letter | example) where Phat is the neural net output of predicted probability for a letter. The
