AIPID: MAD-ML-Powered AIP Discovery Platform

Abstract Inflammation is a biological defense mechanism against harmful stimuli such as infection, tissue injury, or toxic agents, which, if prolonged, can lead to chronic inflammatory disorders. The limited safety and tolerability of current anti-inflammatory drugs emphasize the need for novel, selective therapeutic agents. Anti-inflammatory peptides (AIPs) have emerged as promising candidates owing to their ability to selectively target diseased cells while sparing healthy tissue. However, the identification of AIPs remains constrained by labor-intensive and expensive wet-laboratory screenings. To address this challenge, we have developed AIPID, an interactive, publicly accessible web-application for faster identification of anti-inflammatory peptides. Central to our platform is our Motif-Analysis-Driven Machine Learning (MAD-ML) model, based on the approach of representative negative dataset selection, combining iterative random sampling with motif profiling to enhance dataset diversity and model robustness. Finally, AIPID employs a Random Forest-based classifier trained on motif-filtered, biologically relevant peptide sequences, classifying inputs as AIPs or non-AIPs based on sequence-derived physiochemical descriptors. The model demonstrated excellent performance, achieving sensitivity of 95.16%, specificity of 99.98%, and F1 score of 98.14%, outperforming existing models and correctly predicting 18 of 19 experimentally validated AIPs. The AIPID platform offers an intuitive, multi-page interface for sequence-based peptide prediction, exploration of UniProt-derived AIP repositories, and access to statistical insights on peptide properties. The application is freely available at https://aipid-app-version1.streamlit.app/, providing a valuable resource for the peptide therapeutics research community. Competing Interest Statement The authors have declared no competing interest.