AI-Generated Virtual Libraries of Anti-Inflammatory Phytochemicals and Their Pedagogical Application in Cell Biology, Biochemistry, and Food Chemistry

Abstract

We developed and evaluated an AI‐powered workflow to build a virtual library of anti‐inflammatory phytochemicals for educational use. Text mining of PubMed and Scopus identified 150 candidate compounds, 132 of which were curated and converted into standardized SMILES. A Random Forest QSAR model achieved R² = 0.82 for COX‐2 IC₅₀ prediction, and docking with AutoDock Vina confirmed high binding affinities (−9.2 to −8.0 kcal/mol). The resulting MySQL‐driven web platform allowed undergraduate students to perform structure–activity analyses and molecular docking in class. A post‐module survey (n = 42) showed a significant gain in computational confidence (mean = 2.2; p < 0.001). This approach enhances both research efficiency and computational training in life‐science education.