PM-BioPred: A Web-Server for Prediction of Compound Bioactivity Against Plant and Microbial Proteins

Understanding protein-ligand interactions in plants and microbes is essential for advancing agricultural biotechnology and developing antimicrobial strategies. In plants, these interactions govern critical physiological processes such as growth, immunity, and stress response, while in microbes, they influence pathogenicity, virulence, and survival. Experimental determination of bioactive compound-protein interactions is labor-intensive and limited in scope, and existing computational methods often suffer from rigid modeling assumptions. In this study, we collected and curated experimentally validated active and inactive compounds against plant and microbial target proteins from publicly available databases. Using this curated dataset, we developed machine learning-based classification models that predict compound bioactivity with an accuracy of 86% for plant proteins and over 90% for microbial proteins. These models form the core of PM-BioPred, a web-accessible prediction server designed to assist researchers in identifying potentially bioactive compounds against plant and microbial targets. PM-BioPred provides a user-friendly interface for submitting compound queries and retrieving bioactivity predictions. PM-BioPred is freely accessible to the academic community at https://pmbiopred.streamlit.app/ and serves as a valuable resource for plant-pathogen interaction research and compound repurposing efforts. The platform aims to facilitate the early-stage screening of agrochemicals and antimicrobial candidates.