Explainable QSAR models of 5-HT1A receptor ligands using conceptual DFT descriptors and no-code machine learning...

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Molecular informatics

Background:

• Quantitative Structure-Activity Relationship (QSAR) modeling is crucial for drug discovery.
• Understanding structure-activity relationships aids in designing effective ligands.
• Serotonin 5-HT1A receptor ligands are important therapeutic targets.

Purpose of the Study:

• To develop mechanistically interpretable QSAR models for 89 serotonin 5-HT1A receptor ligands.
• To integrate Conceptual Density Functional Theory (CDFT) descriptors with no-code machine learning.
• To explore the influence of protonation and solvation on ligand activity classification.

Main Methods:

• Computed electronic reactivity indices (electrophilicity, hardness) using CDFT at the GFN1-xTB level.
• Considered neutral and protonated molecular forms in vacuum and aqueous conditions.
• Employed the RandomTree algorithm within an OECD-consistent framework for model construction.

Main Results:

• The electrophilicity index of protonated molecules in aqueous solvent was the primary descriptor.
• Chemical hardness of neutral molecules in the aqueous phase was a secondary descriptor.
• External validation showed moderate but consistent classification performance (Cohen's Kappa 0.49).

Conclusions:

• Accounting for protonation and solvation enhances mechanistic interpretability of CDFT-based QSAR models.
• The no-code workflow offers a transparent and reproducible framework for ligand-based drug discovery.
• This approach facilitates exploratory and mechanistically interpretable molecular informatics.