From Virtual to Real Screening for D₃ Dopamine Receptor Ligands

Abstract

Even in the absence of receptor-structure information, iterative virtual screening cycles with support vector machines (SVM) offer a rapid way to identify novel leads with minimal experimental effort. First, an SVM is trained for prediction of D3 receptor-selective ligands. Based on the prediction of this virtual filter, compounds are tested for binding affinity at D2 and D3 receptors. Second, a similarity search is performed with the most promising candidate from round one. Four out of five compounds from the final hit list exhibited nanomolar affinity at the D3 receptor including a novel scaffold structure. The Ki value of the best molecule (1) was 40±6 nM.