Application of artificial neural networks for predicting imidazole derivatives antimicrobial activity against Enterococcus faecalis

Abstract

Abstract The paper discusses artificial neural networks (ANNs) as a useful deep learning (DL) method to assist researchers in the search for new therapeutic and disinfectant substances. Two ANN models have been designed to predict the biological activity of the compounds based on their physicochemical properties and their structure. The said activity was tested against Enterococcus faecalis bacteria on a series of 140 imidazole derivatives. The regression model designed, predicted the minimum growth inhibitory concentration of E. faecalis (regression model: training data R = 0.91; test data R = 0.91; validation data R = 0.97). The classification model, on the other hand, divided the tested compounds into active or inactive against the tested microorganism predictive (classification accuracy: 92.86%). The exponential demand for new compounds in the pharmaceutical industry, requires alternative experimental methods to reduce the time and cost of development. Therefore, this paper proposes ANN as an alternative to standard techniques for predicting complex biological phenomena.