Diversity and coverage of structural sublibraries selected using the SAGE and SCA algorithms.

Area of Science:

• Computational Chemistry
• Cheminformatics
• Drug Discovery

Background:

• Synthesizing and testing entire chemical libraries is often impractical.
• Rational selection of compound subsets is needed to guide synthetic efforts.
• Diverse and representative library design is key for efficient drug discovery.

Purpose of the Study:

• To compare the performance of Simulating Annealing Guided Evaluation (SAGE) and Stochastic Cluster Analysis (SCA) algorithms.
• To evaluate their ability to select diverse and representative subsets from virtual chemical libraries.
• To assess their utility in directing experimental synthetic efforts.

Main Methods:

• Comparison of two stochastic sampling algorithms: SAGE and SCA.
• Assessment of subset diversity and coverage using u- and s-optimal metrics.
• Testing with simulated 2D datasets and a 27,000-compound proprietary library using Molconn-Z descriptors.

Main Results:

• Both SAGE and SCA algorithms generated diverse and representative sublibraries.
• The simpler SCA method achieved results comparable to the computationally intensive SAGE method.
• Subsets selected by both algorithms demonstrated good coverage of the original library space.

Conclusions:

• SCA is an effective and computationally efficient method for selecting diverse and representative compound subsets.
• The findings support the use of SCA for rational library design and directing synthetic efforts.
• Both algorithms provide valuable tools for optimizing chemical library exploration in drug discovery.