Factors Influencing the Binding of HIV-1 Protease Inhibitors: Insights from Machine Learning Models

Area of Science:

• Structural biology
• Computational chemistry
• Drug discovery

Background:

• HIV-1 protease (PR) inhibitors are vital for acquired immunodeficiency syndrome (AIDS) antiviral therapies.
• Numerous PR-inhibitor complexes are structurally characterized but lack binding affinity data, limiting efficacy assessment.
• A comprehensive understanding of inhibitor efficacy is hindered by this data gap.

Purpose of the Study:

• To develop and validate machine learning models for predicting HIV-1 protease inhibitor binding affinity.
• To identify key molecular features influencing inhibitor efficacy using crystallographic data.
• To bridge the gap in experimental binding affinity measurements for PR-inhibitor complexes.

Main Methods:

• Construction and validation of machine learning models using crystallographic coordinates of 291 PR-inhibitor complexes.
• Leveraging over 2500 molecular descriptors for model training.
• Analysis of three distinct models: KBest with random forest, recursive feature elimination with random forest, and sequential feature selection with support vector machine.

Main Results:

• Machine learning models achieved accuracy scores exceeding 0.85.
• Predicted binding affinities for 274 additional PR-inhibitor complexes lacking experimental data.
• Identified key predictive features including inhibitor charge distribution, hydrogen-bonding, 3D topology, and PR active site symmetry and flap mutations.

Conclusions:

• Machine learning accurately predicts HIV-1 protease inhibitor binding affinity from structural data.
• Structural and chemical properties of both inhibitors and the protease are crucial for efficacy.
• This approach enhances structural understanding and supports the development of novel antiviral therapies.