The enrichment and recall of known inhibitors in a virtual screen are correlated with the probability of finding effective inhibitors through this process. In practice, a large number of false positives are ranked higher than known inhibitors in many virtual screen results. In this paper, we use the interaction of known inhibitors across a range of decoy active sites in order to formulate a modified ranking score, Rscore. This ranking scheme seeks to normalize the DOCK score of a compound based on its interaction with decoy active sites, and uses a linear programming formulation to optimize Rscore for inhibitors versus non-inhibitors. We show an increase in recall of known inhibitors by greater than 20% in most of the test cases considered.