Dale Drueckhammer

We propose to employ the Seawulf cluster to optimize the 3-dimensional structures of virtual molecular databases using quantum calculations. These databases are employed in this group’s molecular design projects using the database-searching program CAVEAT. The two most-used databases called BIAD and TRISUB were previously constructed in this group and contain 2,255 and about 757,000 structures respectively. Currently, the 3-dimensional structures of the molecules in these databases are the results of simple molecular mechanics calculations. For many molecules, the structures calculated by molecular mechanics are not sufficiently accurate for our work. As a result, many false leads are identified in our database searches using CAVEAT that must be weeded out by higher-level calculations on individual structures. Perhaps more importantly, many possible leads are presumably missed in the CAVEAT search because of inaccuracies in the calculated structures of molecules that should be good leads. Our aim is to refine the structures in these virtual databases using Hartree Fock calculations. Initially the structures in the smaller BIAD database will be optimized at the Hartree Fock/3-21G level using the existing structures from molecular mechanics as the input files. This will provide an initial higher-level database for use in this project. Furthermore, this will serve to develop the process and will provide a basis for estimating the computer time required to process the much larger TRISUB database. If it is concluded that Hartree Fock calculations would require so much processor time that this is not feasible, a semiempirical quantum method may be chosen. Such a method would be much faster than the Hartree Fock calculations but would still be expected to provide much more accurate results than the previous molecular mechanics calculations. The larger TRISUB database will be refined in installments and results will be evaluated by analyzing the accuracy of representative structures.