Generation of artificial neural networks models in anticancer study

Artificial neural networks (ANNs) have several applications; one of them is the prediction of biological activity. Here, ANNs were applied to a set of 32 compounds with anticancer activity assayed experimentally against two cancer cell lines (A2780 and T-47D). Using training and test sets, the obtained correlation coefficients between experimental and calculated values of activity, for A2780, were 0.804 and 0.829, respectively, and for T-47D, we got 0.820 for the training set and 0.927 for the test set. Comparing multiple linear regression and ANN models, the latter were better suited in establishing relationships between compounds’ structure and their anticancer activity.