Sustained performance of knowledge-based potentials in fold recognition

We describe the results obtained using fold recognition techniques in our t hird participation in the GASP experiment. The approach relies on knowledge -based potentials for alignment production and fold identification. As indi cated by the increase in alignment quality and fold identification reliabil ity, the predictions improved from CASP1 to CASP3. In particular, we identi fied structural relationships in which no known evolutionary Link exists. O ur predictions are based on single sequences rather than multiple sequence alignments, Additionally, we voluntarily submitted only a single model for each target because, in our view, submission of a single model is the most stringent test, We describe the methods used, the strategy adopted in the p redictions, and the prediction results and discuss future work. (C) 1999 Wi ley-Liss, Inc.