AI-driven in silico discovery, design, and evaluation of novel antiviral candidates targeting Marburg virus nucleoprotein for advanced therapeutic development

Abstract

Background: Marburg virus disease (MVD) is a highly virulent hemorrhagic fever caused by the Marburg virus (MARV), with case fatality rates reaching up to 90 %. Currently, no FDA-approved antiviral drugs exist for MVD, and available treatments are limited to supportive care. This highlights an urgent need for effective and targeted antiviral therapies. Objective: The aim of this study was to design, screen, and evaluate novel antiviral candidates targeting the nucleoprotein of MARV using AI-assisted in silico drug design and molecular docking approaches. Methods: The 3D structure of the MARV nucleoprotein (PDB ID: 5F5M) was retrieved and analyzed to predict active binding sites using computational tools. Crizotinib, an FDA-approved tyrosine kinase inhibitor, was selected as the lead compound based on virtual screening against the AXL gene—a key target associated with MARV infection. Using the WADDAICA AI platform, three novel Crizotinib-derived molecules were generated. Molecular docking was performed via CB-Dock2 to assess binding affinity with the nucleoprotein. Pharmacokinetic and toxicity profiles of the designed molecules were evaluated through ADMET and ProTox-II analyses. Molecular dynamics (MD) simulation was conducted using the iMODS server to validate the stability and interactions of the docked complexes. Results: Among the three designed molecules, Molecule 3 demonstrated the highest binding affinity with the MARV nucleoprotein (− 8.3 kJ/mol). ADMET analysis indicated that Molecule 3 exhibited favorable absorption, distribution, and elimination properties, and was non-mutagenic, non-carcinogenic, and non-hepatotoxic. Toxicity analysis further supported its low-risk profile. Molecular dynamics simulations confirmed stable interactions with low deformability and consistent eigenvalues, validating its structural compatibility as a drug candidate. Conclusion: This AI-driven in silico study successfully identified a promising Crizotinib-derived molecule (Molecule 3) with high binding affinity, favorable pharmacokinetic properties, and minimal toxicity. These findings support the potential of Molecule 3 as a lead antiviral candidate against MVD, warranting further in vitro and in vivo validation to assess its therapeutic efficacy.