A hierarchical algorithm for molecular similarity (H-FORMS).

Area of Science:

• Computational chemistry
• Cheminformatics
• Structural bioinformatics

Background:

• Accurate molecular similarity assessment is crucial for drug discovery and materials science.
• Existing methods often struggle with computational efficiency and scalability for large datasets.
• Identifying molecules with identical structures requires robust alignment and atom matching algorithms.

Purpose of the Study:

• To introduce a new hierarchical method for determining molecular similarity.
• To develop an algorithm capable of detecting identical molecular structures through rigid transformation and atom correspondence.
• To improve upon existing methods in terms of speed and accuracy.

Main Methods:

• A hierarchical approach comparing global spatial structures first.
• Computation of novel, local, rotation-invariant descriptors for atom neighborhoods.
• Application of deterministic and stochastic strategies for alignment computation and atom matching.
• Utilizing the Hungarian algorithm for assignment problems.

Main Results:

• The hierarchical method effectively compares global and local molecular features.
• Local similarity indexes reduce testing trials and the dimensionality of the Hungarian assignment problem.
• The proposed method demonstrates superior performance over state-of-the-art techniques.
• Significant improvements in computational time and accuracy were observed.

Conclusions:

• The new hierarchical method provides an efficient and accurate way to determine molecular similarity.
• The integration of local descriptors enhances the robustness of atom matching.
• This approach offers a promising alternative for large-scale molecular structure comparisons.