SpaceGrow: efficient shape-based virtual screening of billion-sized combinatorial fragment spaces.

Area of Science:

• Cheminformatics
• Computational Chemistry
• Drug Discovery

Background:

• Ultra-large chemical libraries (billions/trillions of compounds) challenge traditional cheminformatics due to size.
• Exhaustive enumeration is infeasible for these massive chemical spaces.
• Existing 3D virtual screening methods often rely on computationally expensive exhaustive enumeration.

Purpose of the Study:

• Introduce SpaceGrow, a novel shape-based 3D approach for virtual screening.
• Enable efficient screening of ultra-large chemical spaces (billions of compounds).
• Improve upon existing 3D virtual screening methods in terms of speed and performance.

Main Methods:

• Developed SpaceGrow, a combinatorial, shape-based 3D virtual screening method.
• Utilized a single CPU for screening billions of compounds within hours.
• Assessed pose reproduction using Root Mean Square Deviation (RMSD) and ranking performance.
• Evaluated SpaceGrow on subsets of the eXplore chemical space and in drug discovery workflows involving G protein-coupled receptors (GPCRs).

Main Results:

• SpaceGrow demonstrates comparable pose reproduction to conventional superposition tools.
• SpaceGrow exhibits superior ranking performance and is orders of magnitude faster.
• Larger chemical spaces show a higher probability of identifying superior results.
• Successful application in identifying novel compounds with similar binding capabilities for GPCRs.

Conclusions:

• SpaceGrow offers an efficient and effective solution for ligand-based virtual screening of ultra-large chemical spaces.
• The method significantly accelerates the drug discovery process by enabling rapid exploration of vast compound libraries.
• Findings highlight the potential of searching ultra-large chemical spaces for drug discovery and development.