An intersection inequality sharper than the tanimoto triangle inequality for efficiently searching large databases

Area of Science:

• Computational chemistry
• Cheminformatics

Background:

• Efficiently searching large molecular databases is crucial for drug discovery and materials science.
• Molecular fingerprints and similarity measures like Tanimoto similarity are widely used for this purpose.
• Existing bounds based on triangle inequality can limit search efficiency.

Purpose of the Study:

• To derive a novel intersection inequality for bounding Tanimoto similarity between molecular fingerprints.
• To demonstrate the superiority of this new bound compared to existing methods.
• To introduce an integer representation for fingerprints and generalize the inequality for efficient database searching.

Main Methods:

• Derivation of a new intersection inequality for binary fingerprint vectors.
• Comparison of the new bound with the bound derived from the triangle inequality of Tanimoto distance.
• Introduction of a generalized integer representation of fingerprints.
• Application of the generalized intersection inequality to the integer representation.

Main Results:

• A new intersection inequality provides a significantly sharper bound on Tanimoto similarity.
• The derived bound is more effective than the bound from the triangle inequality.
• The generalized inequality effectively applies to the novel integer representation of fingerprints.
• The method enables efficient searching of large binary molecular fingerprint databases.

Conclusions:

• The new intersection inequality offers a more precise and efficient method for bounding molecular similarity.
• The integer representation and generalized inequality facilitate faster and more effective database searches.
• This approach has significant implications for accelerating the discovery of novel molecules.