Exploring the functional food potential of Zea mays using machine learning–based QSAR and network biology to identify anti-diabetic and anti-inflammatory phytochemicals

Abstract

Traditional medicinal plants such as Zea mays (maize) display promising anti-diabetic and anti-inflammatory effects, yet their molecular mechanisms remain unclear. This study investigated the structure–activity relationships of Z. mays phytochemicals using a polypharmacology framework. Network pharmacology analyses, including protein–protein interaction mapping, KEGG, Kyoto Encyclopedia of Genes and Genomes, pathway enrichment, and gene annotation, identified AKT1 as the central mediator of the observed effects. A machine learning–guided Quantitative Structure–Activity Relationship (QSAR) model was constructed using PubChem substructure fingerprints and deployed as a predictive web application (https://akt1-pred.streamlit.app/). The model demonstrated strong performance, validated by receiver operating characteristic analysis and applicability domain assessment. Molecular docking and dynamics simulations further supported the predicted binding interactions. Delta-tocopherol, thiamine, and 9-ribosyl-trans-zeatin emerged as the most potent phytochemicals, in some cases showing higher predicted activity than Food and Drug Administration (FDA) approved drugs. These findings provide a mechanistic basis for Z. mays bioactivity, highlight the therapeutic relevance of its phytochemicals, and offer a computational tool to accelerate natural product–based drug discovery.