An integrated approach for novel PTP1B inhibitor screening: combining machine learning models, molecular docking, molecular and dynamics simulations.

Diabetes mellitus, particularly type 2 diabetes (T2DM), is a major global health challenge characterized by persistent hyperglycemia resulting from insulin resistance. Protein tyrosine phosphatase 1B (PTP1B) has emerged as a key enzyme involved in regulating insulin signaling, making it a promising target for therapeutic interventions aimed at improving insulin sensitivity. However, the development of effective PTP1B inhibitors has been hindered by issues such as poor bioavailability and off-target effects. This study presents an integrated approach combining machine learning (ML), molecular docking, and molecular dynamics (MD) simulations to identify novel PTP1B inhibitors. An ML-based predictive model was developed using a dataset of over 2183 known PTP1B inhibitors to guide the selection of compounds with high inhibitory potential. Molecular docking was applied to a compound database of 1.6 million molecules, identifying 1057 promising candidates, which were then refined using the ML model to select the top five compounds. Additionally, the same strategy was applied to a natural product-derived compound database containing 160,000 molecules, leading to the identification of two additional PTP1B inhibitors. This comprehensive approach, combining ML with computational predictions, accelerates the drug discovery process and enhances the reliability of the findings, offering a promising pathway for the development of novel treatments for T2DM and related metabolic disorders.