One of the most important unsolved problems of computational biology is the inability to predict the three-dimensional structure of a protein from its amino acid sequence. In practice, the solution to the protein folding problem demands that two interrelated problems be simultaneously addressed. Potentials that recognize the native state from the myriad of misfolded conformations partly surmounting both problems. A means of secondary and tertiary restraint information to funnel the molecule towards native-like regions. However, such approaches typically generate two to three low energy topologies. Thus, we propose to develop improved protocols to predict secondary structure and tertiary restraints from multiple sequence information. Furthermore, since native state topology generation and section is also crucially dependent on the non-restraint, empirical contributions to the potential, these terms must also be improved. In particular, side chain burial will be more adequately described and local sequence alignments will be employed to develop much more sensitive pair potentials. Furthermore, once low energy topologies are generated, self-consistent tertiary restraints will be derived so that less distorted native-like conformations will be generated. This should enhance the energetic selectivity for native-like as well as by developing computationally more efficient reduced protein sampling techniques as well as by developing computationally more efficient reduced protein models. To establish the range of validity of this approach to tertiary structure prediction, application will be made to large number of sequences of known as well as unknown structure. Significant, independent testing of this algorithm will be done by participating in blind prediction, contests, including CASP3, by making blind predictions of other proteins, and by disseminating all software to other investigators over the Internet.
Showing the most recent 10 out of 121 publications
