Our efforts deal with all the components of structure-based design: methods to extract the maximum structural information with advanced algorithms; computer graphics routines to display structures; refinement procedures; interactive energy minimization with state-of-the-art potential functions; drug design tools such as DOCK and BUILDER. We are just beginning a program to couple organic synthesis with computational searching. Two problems of immediate interest are the development of a synthetic engine, called DIVERSIFY, that uses organic transformations to generate a large family of products from specified starting points; and the development of methods to explore the very large virtual libraries used in combinatorial chemistry. The DOCK software package for searching data bases of organic and metallo-organic compounds for putative ligands has been a central aspect of UCSF drug design efforts over the past ten years. This suite of programs has wide-ranging applications to enzyme systems, nucleic acids, and, in general, problems involving molecular recognition. The current version of the program uses the AMBER intermolecular potential functions to evaluate proposed binding geometries. DOCK was developed to propose novel lead compounds. It has been successful in this goal in a wide variety of systems, including recent, unpublished work identifying novel inhibitors of the HIV reverse transcriptase. Progress has also been made on the software front. A total revision of the BUILDER program has been written to allow fragments of molecules to be joined together (Roe et al, 1995). Atom types, bond lengths, bond angles, and dihedral angles are all checked. Breadth-first and depth-first searches are employed. Usually, several linkages are found that meet the user-set tolerances. The results can be viewed directly in MidasPlus using delegate routines developed by CGL, preferred links can be selected, and the newly constructed molecule can be optimized using interactive or batch techniques. A first step at a graphical user interface has been prepared (M. Young, unpublished). Our laboratory has long been interested in automating NMR assignments. The specific task has been continually redefined as NMR experiments evolved from two dimensional proton data to three and four dimensional heteronuclear efforts. A successful assignment scheme must have built-in flexibility to handle new experiments as they are devised. This requires a high level of data abstraction. Other crucial elements are the user interface, data display and processing, and combining structural and sequential information with the NMR experiments. Our efforts at this general problem have been combined in a program called SPARKY written by Dr. Don Kneller. SPARKY currently runs on Unix-based workstations, with a development and production system on Sun hardware and a prototype system planned for Silicon Graphics machines. SPARKY includes many of the design features discussed above. It integrates data processing, provides a variety of display tools, establishes a hierarchical data base that combines information from primary and secondary structure of biomacromolecules with NMR data. The program is written in C++, a high-level object-oriented language. Dr. Kneller continues to collaborate on the port to the X/Motiv windowing environment have begun. In the past year, the program has been fully documented and and many features have been upgraded.
Showing the most recent 10 out of 508 publications
