Combining experimental and AI approaches to identify antifungal compounds of Origanum vulgare and Mentha pulegium essential oils targeting Candida albicans

Abstract

The growing issue of antifungal resistance and the persistence of microbial biofilms underscores the need for novel therapeutic strategies. This study presents an Artificial Intelligence (AI)-based framework to predict the antifungal and antibiofilm effectiveness of essential oils from Origanum vulgare L. and Mentha pulegium L., both individually and in combination with nystatin, against Candida albicans. Three machine learning models—linear regression, AdaBoost, and Random Forest—were employed to predict the Minimum Inhibitory Concentrations (MICs) and antibiofilm activity. Furthermore, a feature selection algorithm was used to identify the most influential bioactive compounds, with thymol and pulegone emerging as key predictors in the models. Models'performance was assessed using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), and validated through experimental methods, including disk diffusion, microdilution method, antibiofilm efficacy, and the checkerboard method to assess synergistic effects with nystatin. Experimental results closely corroborated the predictions of the proposed AI models, further substantiating the significant antifungal and antibiofilm properties of these essential oils. GC-MS/GC-FID analysis for the identification of active compounds revealed that thymol (53.23 %) and pulegone (70.50 %) were the major constituents of O. vulgare and M. pulegium, respectively. Moreover, the combination of nystatin with O. vulgare essential oil showed synergistic activity against both planktonic and sessile forms of C. albicans. These findings validate the predictive power of AI models, highlighting the potential of integrating data science with biology to optimize personalized antifungal treatments and discover new compounds with high therapeutic potential in combating resistant fungal infections.