Hypershape Recognition: A General Framework for Moment-Based Molecular Similarity

Area of Science:

• Computational Chemistry
• Cheminformatics
• Molecular Modeling

Background:

• Existing molecular similarity methods struggle with inorganic and organometallic compounds.
• Drug design heavily relies on molecular similarity but lacks tools for broader chemical spaces.

Purpose of the Study:

• Introduce Hypershape Recognition (HSR), a novel framework for moment-based similarity assessment.
• Extend molecular similarity calculations to diverse chemical entities beyond organic molecules.
• Develop a versatile tool for comparing 3D chemical representations with atomic features.

Main Methods:

• Developed Hypershape Recognition (HSR), a moment-based similarity assessment framework.
• Represented molecules as N-dimensional 'hypershapes' using atomic coordinates and features (atomic number, isotope, formal charge).
• Utilized a principal component analysis (PCA)-based reference system for hypershape comparison, ensuring continuity across geometry fluctuations and enabling enantiomer distinction.

Main Results:

• HSR successfully calculates similarity scores for inorganic and organometallic compounds, overcoming limitations of existing methods.
• The framework is adaptable for specific applications, such as distinguishing isotopologues and transition-metal complexes with varying oxidation states.
• HSR provides continuous similarity scores and can differentiate enantiomers due to its PC-based reference system.

Conclusions:

• Hypershape Recognition (HSR) provides a versatile and robust framework for molecular similarity assessment across diverse chemical structures.
• The method addresses limitations in current approaches, enabling broader applications in chemical research and development.
• HSR is available as open-source software, promoting accessibility and further development in the field.