Screening for New Antidepressant Leads of Multiple Activities by Support Vector Machines

Abstract

Virtual screening was carried out against 21 biological targets related to depression by support vector machine classification using the same atom-type descriptors. The models were effective as 0.2-0.8 of theoretical enrichments of the external test data sets could be achieved, depending on the target. The set of predicted active molecules had large diversity and contained examples with high dissimilarity to the compounds of training sets. Filtering the database of known antidepressants by all 21 models it was found that on average compounds were classified active for 2.3 targets.