ComfyQSAR reproducibility data for six QSARDB benchmarks and a JAK2 case study

Overview This record contains reproducibility data and analysis outputs generated using ComfyQSAR, a node-based platform for QSAR model development, validation, and external compound screening. Contents The package includes: - Six QSARDB benchmark datasets: QDB258, QDB186, QDB116, QDB260, QDB261, and QDB203- Input structures and descriptor matrices- Preprocessing and descriptor-selection outputs- Selected descriptor sets and trained models- Model evaluation and applicability domain results- JAK2 model-development and external-screening outputs- Prioritized JAK2 candidates and docking results- File inventory and integrity information Software The ComfyQSAR source code is available at: https://github.com/DDUKHAE/ComfyQSAR File integrity `MANIFEST.csv` provides the file path, size, and SHA-256 checksum for each file in the archive. The package also includes information on the software environment used to generate the results. Scope The complete unfiltered descriptor matrices for the external screening libraries are distributed with the ComfyQSAR GitHub repository and are not duplicated in this record.