Date: 2026-05-08
Initial conference benchmark release.
HODC-AES: A Reliability-Aware Benchmark for Higher-Order Drug-Combination Adverse-Event Signal Prediction in Computational Pharmacovigilance
Status: Conference manuscript under preparation
Asif Elahi, Nur Hossain Mahruf, Raihan Kabir, Anowar Hossen Farvez
- GitHub: https://github.com/asifelahii/HODC-AES
- Zenodo DOI: 10.5281/zenodo.20090199
- Zenodo URL: https://zenodo.org/records/20090199
- CORE-AE v3 positive signal-candidate benchmark summary
- Five generated negative protocols
- Random and temporal evaluation artifacts
- Baseline prediction files
- Paper-ready tables
- Paper-ready figures
- Dataset card
- Benchmark description
- Release-readiness checklist
- SHA256 checksums
- Citation template
Random evaluation gives a more optimistic estimate than temporal evaluation. Temporal testing lowers ranking performance, worsens calibration error, and reduces fixed-budget signal recovery.
HODC-AES is a benchmark for reported pharmacovigilance signal candidates. It does not establish causal drug-drug interactions or clinical safety conclusions.