Predictive model for identifying new CYP19A1 ligands on the KNIME analytical platform

Abstract

The purpose of this study was to create a database of the chemical compounds – ligands of human steroid-hydroxylating cytochrome CYP19A1 (aromatase) in order to build a predictive model. The idea was to create a model on the basis of the machinery learning method such as random forest for two types of ligands – with steroidal (I type) and non-steroidal structure (II type). Two predictive models were built with the help of the KNIME analytical platform. Topological descriptors of the chemical structure were used as training data when building a model that takes into account their correlation between the structure of the molecule and the biological effect. The selection of the feature importance of the descriptors, optimal parameters of random forest and the definition of applicability domain of the models were carried out. The assessment of the ability to predict the results of a test sample was performed for each model. The quality marks of the obtained models indicated a rather high predictive ability of the models and the prospects of their use for identification of new human CYP19A1 ligands as potential drugs for treatment of hormone-dependent tumors.