A Workflow Combining Machine Learning with Molecular Simulations Uncovers Potential Dual-Target Inhibitors against...

Area of Science:

• Computational chemistry
• Drug discovery
• Pharmacology

Background:

• Traditional drug development faces challenges with low success rates and multifactorial diseases.
• The one drug-one target approach is often insufficient for complex conditions.
• Multitarget drugs offer potential for improved efficacy and reduced adverse reactions.

Purpose of the Study:

• To develop and validate a computational pipeline for identifying multitarget inhibitors.
• To discover novel dual inhibitors for Bruton's tyrosine kinase (BTK) and Janus kinase 3 (JAK3) as a therapeutic strategy for B-cell lymphoma.

Main Methods:

• A pipeline combining machine learning, SHapley Additive exPlanation (SHAP), and molecular dynamics simulations was employed.
• Natural product databases were screened for potential dual inhibitors.
• Inhibitor candidates were evaluated for drug absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties.

Main Results:

• The pipeline successfully identified prospective dual inhibitors for BTK and JAK3.
• Three candidate inhibitors with acceptable ADMET properties were screened.
• Compound CNP0266747 was selected as the optimal choice, demonstrating favorable binding free energy and specific conformations against BTK and JAK3.

Conclusions:

• The developed computational pipeline is effective for discovering multitarget inhibitors.
• CNP0266747 shows significant potential as a dual inhibitor for B-cell lymphoma treatment.
• Key residues and molecular features contributing to the inhibition of BTK and JAK3 were identified.