PyMolGen: Database-Driven Molecular Generation of Drug-Like Compounds

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Drug discovery programs require controlled generation of novel molecules.
• Existing machine learning methods offer limited user control over molecular generation.
• 3,5-dimethyl-4-phenylisoxazole derivatives are of interest for bromodomain-containing protein 4 inhibition.

Purpose of the Study:

• To present a rule-based approach for de novo molecule generation.
• To provide users with greater control over the generated molecules compared to machine learning methods.
• To apply the method for generating derivatives of a 3,5-dimethyl-4-phenylisoxazole scaffold.

Main Methods:

• Deriving combination rules between molecular fragments from a user-provided database.
• Utilizing a build probability score for each generated molecule.
• Applying the method to generate derivatives of a 3,5-dimethyl-4-phenylisoxazole parent structure.

Main Results:

• Generated molecules are statistically similar to the source database.
• The method allows for constrained optimization at defined vectors.
• The approach provides enhanced user control in molecule generation.

Conclusions:

• The rule-based approach offers a controllable alternative to machine learning for de novo molecule generation.
• This method facilitates the development of targeted drug candidates, such as bromodomain-containing protein 4 inhibitors.
• The build probability score aids in assessing the relevance of generated molecules.