Kevin Gilbert Report
In support for the Big Red Demonstration I developed batch codes to convert the Smiles strings (downloaded from PubChem) to three dimensional coordinates and then export the coordinates in SDF format. The SDF files were then optimized using molecular mechanics calculations and the MMFF94 force field. In addition I wrote a small program to take the SDF formatted structures and convert them to PovRay input files which could then be used to generate pretty pictures for the Big Red Demo. Since then I have rewritten the Smiles parser to be more efficient, and I have tracked down and fixed those structures which were originally problems. I have currently processed 50 of the 512 files that were generated from the original eight million Smiles strings. These files contain approximately 750,000 optimized structures and will be available for processing with Gamess as soon as Mookie Baik is ready for them.
Before proceeding with further calculations, Rajarishi Guha and I have been discussing the format for a simple database to collect and manage all of the structures. At this time we have no method for determining which Smiles strings are unique and whether they have been converted to 3D coordinates already. The original PubChem Smiles string file did not contain PubChem ID's and thus we cannot, at this time, correlate the computed structures with any PubChem information. Rajarishi has now recreated the PubChem Smiles string file with the appropriate PubChem ID information and we will repeat the Smiles to SDF conversion and geometry optimization. At this time it would be appropriate to run these computations on Big Red. We will also run some small scale jobs to generate sufficient data to aid in designing and building a database of structures.