Finding Balance: Multiobjective Optimization in Molecular Generative Modeling.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Developing novel therapeutics requires balancing multiple properties like potency, safety, and metabolic stability.
• Designing compounds for multi-target engagement presents additional complexity.
• Computational methods are crucial for navigating vast chemical space efficiently.

Purpose of the Study:

• To apply multiobjective optimization and generative models for designing novel small molecules.
• To optimize compounds for conflicting pharmacological attributes and potential multi-target interactions.
• To demonstrate a practical strategy for de novo drug design with complex requirements.

Main Methods:

• Utilizing multiobjective optimization techniques combined with generative models.
• Employing predictive modeling to assess compound properties.
• Training models on limited public data for de novo compound generation.

Main Results:

• Successfully generated de novo compounds with a favorable balance of desired properties.
• Demonstrated potential for identified compounds to exhibit affinity for multiple targets.
• Validated the approach across three distinct case studies.

Conclusions:

• The developed computational approach is effective for designing novel small molecules with complex, multiobjective profiles.
• This strategy aids in identifying drug candidates that satisfy conflicting pharmacological requirements.
• The method offers a practical solution for applied drug design challenges.