Estudos in silico na busca de alcaloides inibidores do Fusarium oxysporum

Abstract

The fungus Fusarium oxysporum is a widely distributed phytopathogen responsible for significant agricultural losses and capable of acting as an opportunistic agent in human infections. Control methods commonly employed include the application of fungicides and crop rotation; however, the microorganism has shown resistance mechanisms to different antifungals, limiting the effectiveness of available treatments. In this context, the search for new substances with inhibitory properties becomes essential, aiming at the development of therapeutic alternatives and the improvement of strategies to combat the pathogen. This study aims to identify alkaloids with inhibitory potential against F. oxysporum through an in silico approach. A database from CHEMBL was selected, containing structures previously evaluated in minimum inhibitory concentration (MIC) assays. Subsequently, alkaloid-like structures were analyzed using genetic algorithm and random forest tools; 6 molecular descriptors were selected for class AB. The developed classification model presented consistent statistical parameters in Cohen's kappa values (internal = 0.935; cross = 0.751; external = 0.740), indicating predictability and robustness. Subsequently, the models were evaluated through statistical validation, including the confusion matrix, class statistics (recall, precision, sensitivity, and specificity), general statistics (prediction and overall error, Cohen's kappa, correctly and incorrectly classified structures), and ROC curve. Following virtual screening, structures of 70 active alkaloids with antifungal potential were selected based on 85% reliability. The results highlight the applicability of chemoinformatics in supporting experimental validation and the study of new bioactive structures.