Analysis of proteasome inhibition prediction using atom-based quadratic indices enhanced by machine learning classification techniques

In this work the use of 2D atom-based quadratic indices is shown in the prediction of proteasome inhibition. Machine learning approaches such as support vector machine, artificial neural network, random forest and k-nearest neighbor were used as main techniques to carry out two quantitative structure-activity relationship (QSAR) studies. First, a database consisting of active and non-active classes was predicted with model performances above 85% and 80% in learning and test series, respectively. Second a regression-based model was developed which allow to estimate the EC₅₀ with Q₂ values of 52.89 and 50.19, in training and prediction sets, respectively, were developed. These results provided new approaches on proteasome inhibitor identification encouraged by virtual screenings procedures.