Machine-Learning Approach to Identify Potential Dengue Virus Protease Inhibitors: A Computational Perspective.

Area of Science:

• Virology
• Drug Discovery
• Computational Chemistry

Background:

• Dengue virus (DENV) poses a significant global health threat due to its increasing prevalence and lack of approved antiviral treatments.
• The DENV NS3/NS2B serine protease is crucial for viral replication, making it a key target for antiviral drug development.

Purpose of the Study:

• To identify novel inhibitors of the DENV NS3/NS2B serine protease using machine learning and virtual screening.
• To evaluate the drug-like properties and binding potential of identified inhibitor candidates.

Main Methods:

• Structure-based virtual screening of 32,000 compounds using GNINA.
• Assessment of ADMET properties (absorption, distribution, metabolism, excretion, toxicity).
• Molecular dynamics simulations and MM/GBSA calculations to determine binding free energy.

Main Results:

• Machine learning models successfully screened a large library of potential protease inhibitors.
• Comp530 demonstrated binding potential to the DENV protease as a noncovalent inhibitor.
• Identified compounds offer sites for chemical modification to enhance selectivity and specificity.

Conclusions:

• Machine learning and computational methods can effectively identify potential antiviral drug candidates for DENV.
• Comp530 represents a promising lead compound for developing new dengue antiviral therapies.
• Further optimization of identified compounds may lead to effective covalent or noncovalent inhibitors.