Data-Driven Insights into Porphyrin Geometry: Interpretable AI for Non-Planarity and Aromaticity Analyses

Area of Science:

• Chemistry
• Materials Science
• Computational Chemistry

Background:

• Porphyrins are crucial molecules in various chemical and biological processes.
• Their properties are highly sensitive to subtle structural modifications.
• Machine learning offers a powerful approach to understand structure-activity relationships.

Purpose of the Study:

• To establish a high-quality dataset of metal porphyrins for machine learning analysis.
• To discover and validate structure-property relationships in porphyrins.
• To provide a data-driven foundation for designing tailor-made porphyrins.

Main Methods:

• Curation of 7590 porphyrin structures from the Cambridge crystallographic database.
• Establishment of a dataset comprising 425 metal porphyrins.
• Application of data-driven techniques to analyze nonplanarity and structural aromaticity.

Main Results:

• Validated existing knowledge and discovered new structure-property relationships in porphyrins.
• Demonstrated that distinct nonplanar distortions influence aromaticity differently.
• Found nonplanarity to be more sensitive to macrocycle substitutions than metal or axial ligand effects.

Conclusions:

• Ruffled distortions in porphyrins are primarily influenced by axial ligand size and metal properties.
• These findings offer new insights into tuning porphyrin aromaticity and nonplanarity.
• Machine learning effectively uncovers complex chemical trends in porphyrin structures.