Predicting Drug Combination Activity — Discoverathon 2026 Reproducibility Artifacts

This record contains the large reproducibility artifacts associated with the Discoverathon 2026 project "Predicting Drug Combination Activity". The project develops machine-learning models for predicting anticancer drug-combination activity/synergy using the NCI-ALMANAC dataset. The archive contains: - trained Extra Trees model checkpoints for random and cold-start evaluation;- random train/validation/test splits;- cold-combination train/validation/test splits;- cold-cell-line train/validation/test splits;- cold-drug train/validation/test splits;- the cold-drug assignment file; and- generated test predictions. The source code, documentation, requirements, lightweight results, figures, and reproduction instructions are maintained in the associated public GitHub repository. These artifacts are provided to support reproducibility and independent evaluation of the submitted machine-learning workflow. SHA256 checksum: d506a3b1da38a792145c299eff1fbf70d2a8f6c17a7deb77219cb608c7c0ed37