There are several indices that provide an indication of different types on the performance of QSAR classification models, being the area under a Receiver Operating Characteristic (ROC) curve still the most powerful test to overall assess such performance. All ROC related parameters can be calculated for both the training and test sets, but, nevertheless, neither of them constitutes an absolute indicator of the classification performance by themselves. Moreover, one of the biggest drawbacks is the computing time needed to obtain the area under the ROC curve, which naturally slows down any calculation algorithm. The present study proposes two new parameters based on distances in a ROC curve for the selection of classification models with an appropriate balance in both training and test sets, namely the following: the ROC graph Euclidean distance (ROCED) and the ROC graph Euclidean distance corrected with Fitness Function (FIT(λ)) (ROCFIT). The behavior of these indices was observed through the study on the mutagenicity for four genotoxicity end points of a number of nonaromatic halogenated derivatives. It was found that the ROCED parameter gets a better balance between sensitivity and specificity for both the training and prediction sets than other indices such as the Matthews correlation coefficient, the Wilk's lambda, or parameters like the area under the ROC curve. However, when the ROCED parameter was used, the follow-on linear discriminant models showed the lower statistical significance. But the other parameter, ROCFIT, maintains the ROCED capabilities while improving the significance of the models due to the inclusion of FIT(λ).

A(2B) adenosine receptor antagonists may be beneficial in treating diseases like asthma, diabetes, diabetic retinopathy, and certain cancers. This has stimulated research for the development of potent ligands for this subtype, based on quantitative structure-affinity relationships. In this work, a new ensemble machine learning algorithm is proposed for classification and prediction of the ligand-binding affinity of A(2B) adenosine receptor antagonists. This algorithm is based on the training of different classifier models with multiple training sets (composed of the same compounds but represented by diverse features). The k-nearest neighbor, decision trees, neural networks, and support vector machines were used as single classifiers. To select the base classifiers for combining into the ensemble, several diversity measures were employed. The final multiclassifier prediction results were computed from the output obtained by using a combination of selected base classifiers output, by utilizing different mathematical functions including the following: majority vote, maximum and average probability. In this work, 10-fold cross- and external validation were used. The strategy led to the following results: i) the single classifiers, together with previous features selections, resulted in good overall accuracy, ii) a comparison between single classifiers, and their combinations in the multiclassifier model, showed that using our ensemble gave a better performance than the single classifier model, and iii) our multiclassifier model performed better than the most widely used multiclassifier models in the literature. The results and statistical analysis demonstrated the supremacy of our multiclassifier approach for predicting the affinity of A(2B) adenosine receptor antagonists, and it can be used to develop other QSAR models.