Deep learning approach to the discovery of novel bisbenzazole derivatives for antimicrobial effect

Because of the growing bacterial resistance to antibiotics, the discovery of new antibiotics is critical. The search for new antimicrobial drugs that are effective in treating new and existing microbial diseases is arduous and time-consuming. Deep learning (DL) can help find potential candidates resulting in a more efficient, and cost-effective, and it is more useful on large datasets than other algorithms. Our research team focused on developing an effective DL workflow for discovering new antimicrobial agents. Our group has previously synthesized and tested bisbenzazole structures with various linkers for a variety of pharmacological activities. Antimicrobial activities of bisbenzazole compounds have been also reported in the literature. Deep Neural Networks (DNN) were used to predict the activity of all bisbenzazole compounds synthesized by our group against Staphylococcus aureus and Candida albicans. DNN successfully predicted compounds 16, 17, and 30 out of six molecules (11, 16, 17, 29, 30, and 33) with activity results of 31.25 μg /mL or better results based on in vitro studies. Compounds 13 and 15 out of four molecules (13, 15, 29, and 30) for C. albicans were successfully predicted. Molecular modeling studies were also carried out, and the compounds' docking scores agreed with the DNN models and in vitro antimicrobial activity results. Finally, this workflow, which includes deep learning, molecular docking, and in vitro studies, is a dependable and efficient way of discovering new antimicrobial agents for S. aureus and C. albicans. © 2023