Small-Data QSAR Benchmarking: Impact of Descriptor Dimensionality and Machine Learning Complexity on Predictive Robustness

Initial archived release of the QSAR small-data benchmark study. Includes: Standardized dataset (19 compounds, 29 descriptors) Fully reproducible benchmark pipeline Statistical robustness analysis Y-randomization test LOOCV validation framework This version corresponds to the manuscript submitted to Molecular Informatics. Version update (v2): - Added supplementary metrics table (CSV format).- Added requirements_reproducible.txt for full environment replication.- Updated code (XGBoost n_jobs=1 for deterministic execution).- README aligned with final manuscript title.- Reproducibility verified under Python 3.10.