Machine learning and molecular modeling reveal potential inhibitors of the human metapneumovirus fusion protein.

Abstract

Respiratory infections by human metapneumovirus (HMPV) are common in children, those with weakened immune systems, and older people. With its important role in viral entry, viral fusion (F) glycoprotein is a prime target for designing drugs. To discover new inhibitors of the HMPV fusion protein as a class of drugs that can target this protein and stop it from causing disease, this study employs a computational drug design approach that includes density functional theory (DFT), molecular dynamics (MD), and machine learning (ML). With the help of molecular dynamics simulations, this study verifies the binding activity of lead compounds, optimizes them using calculations based on density functional theory to evaluate electronic properties, and then uses a machine learning-based virtual screening strategy to identify possible inhibitors. PSICHIC, ML model found five lead compounds with ligand 57,414,794 with the highest predicted binding affinity (7.413) and maximum antagonist probability (0.99998). Strong binding of 57,414,794 to the HMPV fusion protein was validated by molecular docking and MM/GBSA binding free energy calculation. The drug outperformed the reference compound Remdesivir with a binding free energy of - 27.46 kcal/mol by a big margin. MD simulations validated its stability with fewer structural fluctuations and good free energy landscape (FEL) characteristics. ADMET profiling also displayed excellent gastrointestinal absorption with no Lipinski violations, supporting the drug-likeness of identified compounds. These results contribute to the search for target-based drugs against HMPV and illustrate the role of machine learning-assisted computational drug design in infectious disease research.