The Jnet secondary structure prediction method. Jnet predicts the secondary
structure of globular proteins from multiple sequence alignments, PSIBLAST profiles and HMMER2 profiles. Jnet also predicts the relative solvent accesibility. Jnet gives 76.4% average accuracy on a large
test set of proteins.
In 2008 JNet was further updated an now gives over 80% accuracy.
You can read the paper that describes the original JNet here.
See the Jnet distribution page to download this program.
JNet is the default prediction method in the JPred server.