Linear to Neural Networks Regression: QSPR of Drugs via Degree-Distance Indices

TL;DR

The paper advances QSPR in drug discovery by linking six VEW-based topological indices, weighted by six atomic properties, to six physicochemical properties across 166 drugs using both linear and nonlinear regressors. It shows Ridge regression often provides robust, interpretable predictions, while nonlinear methods (RF, XGBoost, NN) offer gains for certain properties and reveal index-specific nonlinear interactions. Key indices such as RDD, Harary, TEI, and ECI display strong, property-dependent correlations, suggesting that multi-descriptor VEW representations can enhance rapid property estimation ahead of experiments. These findings highlight the practical potential of VEW topology in cheminformatics to guide early-stage drug design and resource optimization. The work also emphasizes the importance of descriptor selection, model choice, and molecular representation (atomic-number vs atomic-mass weighting) in achieving reliable QSPR predictions.

Abstract

This study conducts a Quantitative Structure Property Relationship (QSPR) analysis to explore the correlation between the physical properties of drug molecules and their topological indices using machine learning techniques. While prior studies in drug design have focused on degree-based topological indices, this work analyzes a dataset of 166 drug molecules by computing degree-distance-based topological indices, incorporating vertex-edge weightings with respect to different six atomic properties (atomic number, atomic radius, atomic mass, density, electronegativity, ionization). Both linear models (Linear Regression, Lasso, and Ridge Regression) and nonlinear approaches (Random Forest, XGBoost, and Neural Networks) were employed to predict molecular properties. The results demonstrate the effectiveness of these indices in predicting specific physicochemical properties and underscore the practical relevance of computational methods in molecular property estimation. The study provides an innovative perspective on integrating topological indices with machine learning to enhance predictive accuracy, highlighting their potential application in drug discovery and development processes. This predictive may also explain that establishing a reliable relationship between topological indices and physical properties enables chemists to gain preliminary insights into molecular behavior before conducting experimental analyses, thereby optimizing resource utilization in cheminformatics research.

Figures (10)

• Figure 1: Heatmap of the correlation between features
• Figure 2: $R^2$, $RMSE$ and $MAE$ scores for different models based on Atomic Mass
• Figure 3: $R^2$, $RMSE$ and $MAE$ scores for different models based on Atomic Number
• Figure 4: Feature Importance Heatmap: Random Forest vs XGBoost with respect to Atomic Mass
• Figure 5: Heatmap of $R^2$ scores for models with respect to Electronegativity