This paper presents an approach to predict the activity of analogues of 2,4,6-trisubstituted 1,3,5-triazines as cannabinoid receptor (CB2) agonists using random forest technique. We compute twenty molecular descriptors for a data set of 58 analogues for the component, and depending on values of these descriptors we train random forest to find a relation between biological activity and molecular structure of analogues. The results obtained by random forest were compared with the decision tree and support vector machine classifiers and the random forest has 100% overall predicting accuracy and for decision tree and support vector machine were 93% and 67% respectively. |
