Artificial Neural Networks Based on CODES Descriptors in Pharmacology: Identification of Novel Trypanocidal Drugs against Chagas Disease

A supervised artificial neural network model has been developed for the accurate prediction of the anti-T. cruzi activity of heterogeneous series of compounds. A representative set of 72 compounds of wide structural diversity was chosen in this study. The definition of the molecules was achieved from an unsupervised neural network using a new methodology, CODES program. This program codifies each molecule into a set of numerical parameters taking into account exclusively its chemical structure. The final model shows high average accuracy of 84 (training performance) and predictability of 77 (external validation performance) for the 4:4:1 architecture net with different training set and external prediction test. This approach using CODES methodology represents a useful tool for the prediction of pharmacological properties. CODES© is available free of charge for academic institutions.