Comparing Molecules Generated by MMPDB and REINVENT4 with Ideas from Drug Discovery Design Teams

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Drug discovery relies on human expertise for designing novel molecules.
• Computational tools offer potential to augment human ideation in early-stage drug discovery.

Purpose of the Study:

• To compare molecules designed by human experts with those generated by MMPDB and REINVENT4.
• To assess the value of computational tools in complementing human expertise in drug discovery.

Main Methods:

• Seven diverse test cases with varied chemotypes were analyzed.
• Molecular structures and properties from human design and computational tools were compared.

Main Results:

• MMPDB and REINVENT4 explored larger chemical spaces than human-designed molecules.
• The chemical spaces explored by MMPDB and REINVENT4 were distinct from each other and from human designs.
• Neither computational tool fully encompassed the chemical space identified by human experts.

Conclusions:

• Computational methods are complementary to human expertise and to each other.
• Generative molecule design tools can accelerate therapeutic candidate identification by expanding explored chemical space.