FLASHFLOOD: a 3D field-based similarity search and alignment method for flexible molecules

Area of Science:

• Computational Chemistry
• Cheminformatics
• Molecular Modeling

Background:

• Flexible molecules present challenges in similarity searching due to their conformational variability.
• Existing methods often struggle with accurately representing and comparing diverse molecular shapes and properties.

Purpose of the Study:

• To develop a 3D field-based similarity search and alignment method for flexible molecules.
• To represent molecular conformational space using fragments and torsional angles.
• To introduce a context-adaptive descriptor scaling for tunable similarity assessment.

Main Methods:

• Representing molecular conformations using fragments and torsional angles.
• Computing fragment pair features using user-definable property fields (generalizing CoMMA descriptors).
• Employing a context-adaptive descriptor scaling procedure for similarity basis.
• Utilizing an assembly algorithm to merge fragment pairs into aligned full structures.

Main Results:

• Demonstrated improved efficiency and accuracy on the dihydrofolate/methotrexate benchmark system by incorporating contextual information.
• Successfully queried a database of approximately 23 million conformers from seventeen flexible molecules.
• Provided computational timings for the search and alignment process.

Conclusions:

• The proposed 3D field-based method offers an efficient and accurate approach for flexible molecule similarity searching and alignment.
• Context-adaptive descriptor scaling is crucial for optimizing performance based on specific research contexts.
• The method is scalable to large molecular databases, showing practical applicability in drug discovery and chemical informatics.