Feature Selection for ADMET Prediction using Genetic Algorithms

SOTO AXEL JUAN; CECCHINI ROCÍO LUJÁN; VAZQUEZ GUSTAVO ESTEBAN; PONZONI IGNACIO

Mar del Plata, Buenos Aires

Congreso; 36 JAIIO: 36º Jornadas Argentinas de Informática e Investigación Operativa - ASAI 2007: IX Argentine Symposium On Artificial Intelligence; 2007

SADIO - Sociedad Argentina de Informática

In this work, a novel approach for descriptor selection aimed to physicochemical property prediction is presented. The capacity of determining the most significant set of descriptors is of great importance due to their contribution for improving ADMET prediction models. The proposed methodology combines a genetic algorithm with decision trees. An experimental analysis was carried out for predicting the octanol-water partition coefficient (logP) using neural networks as prediction method. The performance results showed the good potential of this technique. Other techniques were also tested in order to be compared with the presented proposal.