In silico Compound Screening for Drug Discovery in the “Cloud”

Abstract

The need to identify new small molecules and novel binding partners for known bioactive sites remains a constant in drug development. We are aiding VaSSA Informatics in creating informatics methods used to identify functionally similar compounds ab initio, regardless of structural concerns. To achieve this goal, we implemented an information content algorithm as the primary screening parameter. The results of ChemVaSSA's validation cycle suggest that it can, in fact, detect functionally similar molecules that interact with known bioactive sites ab initio. Validation of ChemVaSSA's results was performed using in silico modeling. First, we developed a ligand library that contained the information content signature of 600,000 ligand/compound complexes for which structure was available. We then utilized atorvastatin (Lipitor) as a test compound to search for molecules with similar functional roles. In addition to returning structurally expected results (other statins), we identified several compounds that, via modeling, appear to bind Hmg-CoA reductase at the site of Lipitor binding but that are NOT structurally similar to Lipitor or other statins. We are structuring this screening to utilize the Amazon EC2 resource and show the cost model associated with this.