Data driven QSPR modeling of psychiatric drugs using degree based graph invariants

Area of Science:

• Computational chemistry
• Cheminformatics
• Medicinal chemistry

Background:

• Molecular descriptors are crucial for predicting drug properties.
• Quantitative Structure-Property Relationship (QSPR) analysis links molecular structure to properties.
• Artificial intelligence (AI) and graph theory offer advanced predictive tools.

Purpose of the Study:

• To evaluate degree-based topological indices in QSPR for drug compounds.
• To compare classical regression models with artificial neural networks (ANNs).
• To identify key topological indices for drug property prediction.

Main Methods:

• Computation of a comprehensive set of topological descriptors.
• Application of linear, polynomial regression, and artificial neural networks (ANNs).
• QSPR analysis of various drug compounds, including psychiatric drugs.

Main Results:

• ANN models demonstrated superior predictive accuracy and robustness over traditional methods.
• Nonlinear relationships between molecular structure and properties were effectively captured by ANNs.
• Specific topological indices were identified as important predictors of drug properties.

Conclusions:

• The combination of topological indices and AI techniques provides an efficient framework for accurate drug property prediction.
• ANNs are highly effective for QSPR analysis, especially for complex, nonlinear molecular properties.
• This approach has significant implications for drug discovery and optimization processes.