Multi-Objective Feature Selection in QSAR Using a Machine Learning Approach

SOTO, AXEL J.; CECCHINI, ROCÍO L.; VAZQUEZ, GUSTAVO E.; PONZONI, IGNACIO

MOLECULAR INFORMATICS

Wiley InterScience

Lugar: Weinheim, Germany; Año: 2009 vol. 28 p. 1509 - 1523

1868-1751

The selection of descriptor subsets for QSAR/QSPR is a hard combinatorial problem thatrequires the evaluation of complex relationships in order to assess the relevance of theselected subsets. In this paper, we describe the main issues in applying descriptor selectionfor QSAR methods and propose a novel two-phase methodology for this task. The firstphase makes use of a multi-objective evolutionary technique which yields interestingadvantages compared to mono-objective methods. The second phase complements thefirst one and it enables to refine and improve the confidence in the chosen subsets ofdescriptors. This methodology allows the selection of subsets when a large number ofdescriptors are involved and it is also suitable for linear and nonlinear QSAR/QSPRmodels. The proposed method was tested using three data sets with experimental valuesfor blood-brain barrier penetration, human intestinal absorption and hydrophobicity.Results reveal the capability of the method for achieving subsets of descriptors with ahigh predictive capacity and a low cardinality. Therefore, our proposal constitutes a newpromising technique helpful for the development of QSAR/QSPR models.