A method for protein secondary structure prediction
The 513 set contains proteins that are non redundant to 5SD, and as such they can be used for training new secondary structure prediction methods.
The 126 set of proteins was only run cross-validated for PHD. As such the predictions do not represent current accuracies. For a full review see Cuff J. A. and Barton G. J. Proteins (1998).
The links allowing sorting will no longer work, this may change if you ask nicely.
If you have rasmol setup to accept MIME-type scripts you can also see the structure in 3D.