Interpretable Quantitative Structure–Activity Relationship (QSAR) for identification of potent antifungal activity agents towards Candida albicans ATCC 2091

Fungal infections are an increasing global health issue. Despite available treatments, fungal resistance reduces medicine effectiveness. This research conducted QSAR analysis on fifty-one 4-aryl-2-hydrazinothiazole derivatives previously evaluated for antifungal activity. The QSAR model was derived from a hybrid method combining genetic algorithms (GA) and multiple linear regression (MLR). The analysis showed a negative correlation between pMIC and RDF100e, ITH, R4m+, RDF120s, and GATS8e. The model was validated using an external test set by the leave-one-out cross-validation method. Additionally, Y-randomization, MAE, and Golbraikh-Tropsha metrics assessed the model's applicability domain. The study offers an in-depth molecular descriptor interpretation through three methods: atomic pair distribution, substructure-based analysis, and molecular surface mapping with cumulative atomic contributions. These methods help identify favorable and unfavorable structural groupings. Key molecular features influencing antifungal activity were identified, particularly the spatial arrangement of N1-hydrazine and C4 fragments in the thiazole nucleus. The research highlights Van der Waals interactions, electronegative atoms in substituents, and electron-donating groups. To address the limitations of modeling a small dataset, we applied the novel ARKA approach-based on Arithmetic Residuals in K-groups Analysis-to reduce descriptor dimensionality while preserving chemical relevance and improving interpretability.