A support vector machine (SVM)-based virtual screening method is demonstrated as a rapid computational tool for the prediction of potential asymmetric organocatalysts for the direct aldol reaction. Our models show good accuracy at cross-validation and independent testing. Structure analyses of screening hits from the PubChem database revealed several new classes of compounds. including beta-amino acids. diamines and hydrazides, as potential chiral organocatalysts. (C) 2009 Elsevier B.V. All rights reserved.