QSAR Classification Models for Predicting Affinity to Blood or Liver of Volatile Organic Compounds in e-Health

MARTÍNEZ, MARÍA JIMENA; DIAZ, MÓNICA FÁTIMA; PONZONI, IGNACIO; CRAVERO, FIORELLA

Granada

Conferencia; IWBBIO 2017 (5th International Work-Conference on Bioinformatics and Biomedical Engineering); 2017

In this work, we present Quantitative Structure-Activity Relationship(QSAR) classification models for characterization of molecules affinity to bloodor liver for volatile organic compounds (VOCs), using information providedfrom log Pliver measures for VOCs. The models are computed from a dataset of122 molecules. As a first phase, alternative subsets of relevant moleculardescriptors related to the target property are selected by using feature selectionmethods and visual analytics techniques. From these subsets, several QSARmodels are inferred by different machine learning methods. These models allowclassifying a new compound as a molecule with affinity to blood, to the liver orequal affinity to both. The model with the highest performance correctly classifies72.13% of VOCs and has an average receiver operating characteristic areaequal to 0.83. As a conclusion, this QSAR model can predict the mediumaffinity of a VOC, which can help in the development of physiologically basedpharmacokinetic computational models required in e-health.