Second-generation de novo design: a view from a medicinal chemist perspective

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Retrospective validation of computational de novo design is challenging.
• Existing tools often lack comprehensive validation frameworks.
• NovoBench project integrated multiple programs and expertise.

Purpose of the Study:

• To propose a comprehensive workflow for de novo design.
• To establish a general validation scheme for de novo design techniques.
• To assess the performance of the FlexNovo search engine.

Main Methods:

• Fragment-based de novo ligand reconstruction.
• Utilized the NovoBench project's suite of programs.
• Employed the FlexNovo structure-based de novo search engine.
• Assessed synthetic accessibility using the SYLVIA program.

Main Results:

• FlexNovo reconstructed native ligands in 5 out of 8 cases.
• Reconstructed ligands were ranked within the top 5 candidates in 4 out of 5 successful cases.
• Generated molecules exhibited comparable synthetic complexity to native ligands.
• Plausibility of candidates as active compounds was confirmed via literature searches.

Conclusions:

• The proposed workflow and validation scheme are effective for de novo design.
• FlexNovo demonstrates significant potential for reconstructing native ligands.
• The approach balances design novelty with synthetic feasibility and biological relevance.