Fragment Merging Using a Graph Database Samples Different Catalogue Space than Similarity Search.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Fragment merging is a key strategy for developing potent drug candidates by combining structural motifs of overlapping fragments.
• Identifying suitable fragment merges from commercial databases is challenging but crucial for efficient drug development.
• Synthetic accessibility is a major hurdle in fragment-based drug discovery.

Purpose of the Study:

• To demonstrate the effectiveness of the Fragment Network, a graph database, in identifying fragment merges.
• To compare the Fragment Network approach with traditional fingerprint-based similarity searches for finding fragment merges.
• To showcase the Fragment Network's capability in discovering potent drug candidates.

Main Methods:

• Utilized a Fragment Network database containing over 120 million compounds to identify fragment merges.
• Applied the Fragment Network to four crystallographic screening campaigns.
• Compared results with a traditional fingerprint-based similarity search.

Main Results:

• The Fragment Network identified complementary sets of fragment merges compared to similarity searches.
• Both methods successfully recapitulated fragment-protein interactions but explored different chemical spaces.
• Retrospective analyses identified potential inhibitors with micromolar IC50 values for COVID Moonshot and Mycobacterium tuberculosis EthR.

Conclusions:

• The Fragment Network is a powerful tool for exploring chemical space and identifying fragment merges.
• This graph database approach enhances the yield of fragment merges beyond classical catalogue searches.
• The Fragment Network facilitates the progression of fragments to on-scale potency, aiding drug discovery efforts.