LoFT: similarity-driven multiobjective focused library design.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Combinatorial library design is crucial for drug discovery.
• Existing methods may lack efficiency in optimizing libraries based on multiple criteria.
• Fragment-based approaches offer potential for targeted library generation.

Purpose of the Study:

• To introduce LoFT, a novel tool for focused combinatorial library design.
• To implement a similarity-driven, product-based library design approach at the fragment level.
• To utilize feature tree descriptors for efficient molecular comparison and library diversification.

Main Methods:

• LoFT employs algorithms for constructing focused libraries from chemical fragment spaces.
• A weighted multiobjective scoring function based on physicochemical descriptors guides the search.
• The feature tree descriptor is used for similarity comparisons and library diversification.

Main Results:

• LoFT enables efficient evaluation of chemical fragments without explicit enumeration.
• Validation on three datasets demonstrated optimization of libraries based on similarity to bioactive molecules.
• Performance was compared favorably with existing methods like FTrees-FS.

Conclusions:

• LoFT provides an effective tool for focused combinatorial library design.
• The feature tree descriptor facilitates efficient fragment-level similarity assessment and library diversification.
• LoFT advances computational approaches in drug discovery and medicinal chemistry.