Fast 3D shape screening of large chemical databases through alignment-recycling.

Area of Science:

• Computational chemistry
• Cheminformatics
• Structural bioinformatics

Background:

• Searching large chemical databases for similar structures is crucial.
• Existing graph-based methods are efficient, but 3D shape similarity searches, especially those using 3D shape overlays, remain computationally challenging.
• A new hybrid methodology, alignment-recycling, is proposed to address these challenges.

Purpose of the Study:

• To develop a fast and efficient method for 3D shape similarity searching in large chemical databases.
• To reduce the computational cost associated with 3D shape overlay calculations.
• To enable rapid retrieval and alignment of structures with similar 3D conformations.

Main Methods:

• The alignment-recycling methodology was developed, inspired by techniques for comparing molecular shapes.
• It involves three main steps: selecting diverse shapes, overlaying database conformers to these diverse shapes, and using common reference shapes for query and database conformer overlays.
• Transformation matrices from initial overlays are reused to accelerate subsequent calculations.

Main Results:

• A diverse set of several thousand structures covering the 3D shape space of over one million PubChem compounds was obtained.
• The alignment-recycling method reduced CPU time for shape overlay by 100-fold compared to traditional methods.
• The heuristic approach yielded results comparable to de novo alignment, with over 80% overlap in hit lists on average.

Conclusions:

• Overlay-based 3D similarity searches are computationally intensive for large datasets.
• Alignment-recycling significantly decreases CPU time for these searches by optimizing the alignment process.
• While initial precomputation is required, subsequent database querying becomes two orders of magnitude faster, with potential for extension to larger and more flexible molecules.