MMP-9i Predictor v1.0: an integrative machine learning platform for predicting MMP-9 inhibitors using chemical space analysis and explainable models.

The inhibition of MMP-9 has emerged as a promising therapeutic strategy for treating diseases such as cancer, arthritis, and central nervous system disorders. Despite the potential benefits, the development of MMP-9 inhibitors (MMP-9i) has faced significant challenges, particularly concerning selectivity and off-target toxicity. In this context, we performed a comprehensive cheminformatics study (including chemical space analysis and machine learning approaches) to investigate the structural and physicochemical determinants of MMP-9 inhibitory activity, aiming to facilitate the rational design of promising inhibitors. Model interpretability was enhanced through SHAP analysis, which elucidated the contribution of key features. To improve accessibility for the scientific community, we developed the MMP-9i Predictor v1.0 web tool ( https://mmp-9ipredictor.streamlit.app/ ) to enable prospective screening of novel MMP-9i. It also enables a reliable distinction between inhibitors and non-inhibitors and assesses their applicability domains (AD). Additionally, it visualizes input structures in 2D, enhancing interpretability for researchers.