Toxicity prediction using 2D pharmacophores and support vector machines

Max PereiraAdemar Schmitz

In silico methods have been largely used in drug development process to predict the toxicity of molecules. Predicting the toxicity is one of the most important stage in developing new pharmaceuticals and computational methods are being used in order to make this process less time-consuming and decrease its high cost. Here we report a new approach, using two-dimensional pharmacophore fingerprint to encode pharmacophoric features of molecules in string sets, which are then processed by support vector machines (SVM) to predict the toxicity endpoint of a carcinogenic data set with 1547 compounds. Previous studies have shown the use of machine learning approaches in predicting the toxicity of molecules, however, in those cases it was required to calculate a large number of molecular descriptors to be able to make such prediction. Using SVM and only one molecular descriptor it was possible to achieve a satisfactory accuracy rate compared to other machine learning approaches.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: