Авторы: Сусан Джьякхво, Бочарова В. А., Серов Н.С., Андрей Дмитренко, Виноградов В.В.
Advanced Materials Technologies, DOI: 10.1002/admt.202400458
Abstract
For years, researchers have searched for novel antibiotics to combat pathogenic infections. However, antibiotics lack specificity, harm beneficial microbes, and cause the emergence of antibiotic-resistant strains. This study proposes an innovative approach to selectively eradicate pathogenic bacteria with a minimal effect on non-pathogenic ones by discovering selectively antimicrobial nanoparticles. To achieve this, a comprehensive database is compiled to characterize nanoparticles and their antibacterial activity. Then, CatBoost regression models are trained for predicting minimal concentration (MC) and zone of inhibition (ZOI). The models achieve a ten-fold cross-validation (CV) R2 score of 0.82 and 0.84 with root mean square error (RMSE) of 0.46 and 2.41, respectively. Finally, a machine learning (ML) reinforced genetic algorithm (GA) is developed to identify the best-performing selective antibacterial NPs. As a proof of concept, a selectively antibacterial nanoparticle, CuO, is identified for targeted eradication of a pathogenic bacteria, Staphylococcus aureus. A difference in minimal bactericidal concentration (MBC) of 392.85 µg mL−1 is achieved when compared to non-pathogenic bacteria, Bacillus subtilis. These findings significantly contribute to the emerging research domain of selectively toxic (SelTox) nanoparticles and open the door for future exploration of synergetic interactions of SelTox nanoparticles with drugs.