Novel atomic-level-based AI topological descriptors: application to QSPR/QSAR modeling.

Area of Science:

• Computational Chemistry
• Cheminformatics
• Quantitative Structure-Activity Relationships (QSAR)

Background:

• Existing molecular descriptors struggle to differentiate heteroatoms in complex molecular graphs.
• Accurate modeling of structure-property/activity relationships (QSPR/QSAR) is crucial for drug discovery and materials science.
• The Xu index and Kier-Hall connectivity indices have been foundational but require extensions for broader applicability.

Purpose of the Study:

• To develop novel atomic-level Artificial Intelligence (AI) topological indexes for molecular structure coding.
• To extend the Xu index and AI indexes to compounds containing heteroatoms using a novel vertex degree.
• To build and validate QSPR/QSAR models for physical properties and biological activities using these enhanced indices.

Main Methods:

• Development of AI topological indexes based on graph adjacency and distance matrices.
• Extension of existing indices using a novel vertex degree derived from Kier-Hall's valence connectivity.
• Application of Multiple Linear Regression (MLR) to establish QSPR/QSAR models.

Main Results:

• High-quality QSPR/QSAR models were achieved for physical properties and biological activities of alcohols.
• Molecular size was identified as a dominant factor for physical properties, with minor influences from other atomic types.
• -OH groups significantly impact both physical properties (via hydrogen bonding) and biological activities.

Conclusions:

• The novel AI topological indexes and modified Xu indices are effective parameters for QSPR/QSAR analysis.
• These indices successfully address the challenge of differentiating heteroatoms in molecular graphs.
• The findings underscore the importance of specific functional groups, like -OH, in determining molecular behavior and activity.