Similarity searching in large combinatorial chemistry spaces

Area of Science:

• Computational Chemistry
• Cheminformatics
• Drug Discovery

Background:

• Similarity searching is crucial for navigating large chemical spaces in drug discovery.
• Existing methods often struggle with the scale and complexity of combinatorial chemistry spaces.

Purpose of the Study:

• To introduce Ftrees-FS, a novel algorithm for efficient similarity searching in large chemical spaces.
• To enable the identification of structurally diverse yet similar compounds to a query molecule.

Main Methods:

• Utilizes dynamic programming and a feature tree similarity measure to represent molecules as tree structures.
• Handles combinatorial chemistry spaces holistically, avoiding enumeration of all possible compounds.
• Employs a large dataset of drug fragments (RECAP procedure) as the search space, exceeding 10^18 possible compounds.

Main Results:

• Ftrees-FS identifies the most similar compounds within minutes, even in extremely large chemical spaces.
• The algorithm allows control over the diversity of the generated compound sets.
• Demonstrates successful identification of known active compounds (e.g., dopamine D4, histamine H1, COX2 inhibitors) and ability to bridge structurally unrelated classes.

Conclusions:

• Ftrees-FS offers a powerful and efficient approach for similarity searching in massive chemical libraries.
• The method facilitates the discovery of novel chemical entities with potential therapeutic applications.
• Enables exploration of diverse chemical structures for target-based drug design.