A scalable approach to combinatorial library design for drug discovery

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Combinatorial drug discovery relies on large libraries of diverse and representative compounds.
• Current methods for library design can be computationally intensive and may not account for experimental constraints.

Purpose of the Study:

• To propose a computationally efficient algorithm for designing lead generation libraries.
• To simultaneously optimize compound diversity and representativeness while considering experimental resource limitations.

Main Methods:

• Development of a scalable algorithm leveraging deterministic annealing for cluster identification.
• Truncation of computations from the entire dataset to individual clusters to enhance efficiency.
• Analysis of the trade-off between computational effort and truncation error.

Main Results:

• The algorithm efficiently designs lead generation libraries.
• Demonstrated improvement in computational efficiency by factors of 10 or more on test datasets.
• Validation of the algorithm's performance and analysis through test data applications.

Conclusions:

• The proposed algorithm offers an efficient and effective approach to lead generation library design.
• The method successfully balances diversity, representativeness, and computational cost.
• This framework provides a valuable tool for optimizing drug discovery efforts.