Dataset and Computational Workflow for: Solubility modeling of pazopanib in cosolvent mixtures evaluating thermodynamics versus artificial intelligence
Description
This repository contains the experimental data and the complete computational Python workflows used for the thermodynamic and machine learning modeling of the solubility of the antineoplastic drug pazopanib. The dataset includes equilibrium solubility data in four binary cosolvent systems: (ethanol + acetonitrile), (n-propanol + acetonitrile), (2-propanol + acetonitrile), and (1-butanol + acetonitrile) at temperatures ranging from 288.15 K to 328.15 K.The provided Python scripts (designed for Google Colab) execute a comparative chemometric analysis between classical thermodynamic approaches and artificial intelligence algorithms. The workflow includes: 1) Ideal Solubility Model: Calculations based on the enthalpy and melting temperature of the crystalline lattice. 2) van 't Hoff Semi-empirical Model: Linear regression models to estimate apparent thermodynamic properties of solution (enthalpy and entropy) and phase behavior. 3) Random Forest (RF) Regressor: Machine learning architectures (both generalized and individualized by cosolvent) trained exclusively with extreme pure-solvent data to predict complex non-linear molecular interactions. This dataset aims to facilitate the reproducibility of the statistical metrics (such as R^2, AARD%, RMSE, and MAE), parity plots, and thermodynamic analyses presented in the associated manuscript.
Files
Steps to reproduce
To reproduce the findings and execute the computational models: 1) Download the repository and upload the primary dataset file (PazopanibIA.xlsx / Data.xlsx) to your Google Drive or local working directory. 2) Open the provided Python scripts or Jupyter/Colab Notebooks (.ipynb / .py). 3) Update the directory paths in the "Drive Mounting and Data Loading" code blocks to point to the location where you saved the Excel dataset. 4) Execute the code blocks sequentially. The scripts are structured to automatically generate output folders containing the consolidated statistical reports (.xlsx) and high-resolution parity plots (.png). 5) Required Python libraries: numpy, pandas, matplotlib, seaborn, scikit-learn. Data files and parameters Data.xlsx: The main raw dataset containing experimental solubility values. Columns include Temperature (T in Kelvin), Cosolvent name, Cosolvent mass fraction (w1), and Experimental solubility in mole fraction (x3). Solubility_of_Pazopanib_Colab.ipynb / .pdf: The source code containing the sequential implementation of the Ideal Solubility model, the van 't Hoff model, and the Artificial Intelligence (Random Forest) evaluations. Generated Outputs: Running the scripts will produce diagnostic parity plots (e.g., Evaluation_Plot_vantHoff.png, Evaluacion_Grafica_IAF.png) and tabular statistical summaries comparing the predictive performance of each model.