Degree-Based Prodeg Indices in Graph Theory and Cheminformatics with Sharp Results and QSPR Modeling: Data and Codes

This Zenodo record provides the curated dataset and reproducible code used in the study on degree-based Prodeg descriptors for quantitative structure–property relationship (QSPR) modeling of physicochemical properties. Files in this record Prodeg_QSPR_dataset_Chemeo_2026-01-07_v1.xlsxCurated QSPR dataset for n=90 organic compounds (aromatic-carboxylate series), including molecular identifiers (molecular name, InChI, InChIKey, molecular formula, canonical SMILES), four degree-based Prodeg descriptors (Z−1/2, IP, IP‾, MPI), and physicochemical endpoints compiled from Cheméo. ip-prodeg-index-qspr.ipynbModel benchmarking notebook implementing the full QSPR workflow, including the Kennard–Stone external split (80:20), training-only preprocessing, 5-fold cross-validation and hyperparameter tuning, final refit on the training set, and evaluation on the fixed external test set. Reported metrics include R^2, RMSE, and MAE. Prodeg_Indices_calculation.ipynbThe descriptor computation notebook is used to calculate the Prodeg indices (Z−1/2, IP, IP‾, MPI) from hydrogen-suppressed molecular graphs. Data provenance Physicochemical property values were compiled from Cheméo (https://www.chemeo.com/) on 07 Jan 2026 (units as reported). The curated dataset includes record identifiers, and additional figures/tables are provided in the associated supplementary materials of the study.