Standard Tier Module
Active-Learning Ligand Screening
A research pathway for Bayesian-optimization virtual screening: upload your library (surrogate score, uncertainty, oracle scores), and it drives one step of the drift-aware campaign - Expected-Improvement next-batch selection (downloadable), an escalation recommendation (acquire vs escalate oracle fidelity) via the substrate closure ladder, plus retrospective correlation / top-1% overlap / retrieval / enrichment.
See it run - a worked example, 100% in this browser tab
What it is
A research pathway for Bayesian-optimization virtual screening: upload your library (surrogate score, uncertainty, oracle scores), and it drives one step of the drift-aware campaign - Expected-Improvement next-batch selection (downloadable), an escalation recommendation (acquire vs escalate oracle fidelity) via the substrate closure ladder, plus retrospective correlation / top-1% overlap / retrieval / enrichment. Deterministic; the ML surrogate is your input, not trained here.
Honest scope
Deterministic and citation-backed: every figure is exact arithmetic or a cited rule. Any year- or jurisdiction-indexed value is a confirmable input, never an eternal hardcode. This is a computation tool, not professional (legal, tax, medical, or financial) advice - confirm against the controlling authority for your context.
Authorities cited
- Andersen, Rausch-Dupont, Martinez Leon, Volkamer, Hub, Klakow (2026). Accelerating ligand discovery by combining Bayesian optimization with MMGBSA-based binding affinity calculations. Digital Discovery, DOI 10.1039/d5dd00522a - active-learning pipeline, EI acquisition, top-1% retrieval, Fig 2 correlation/overlap.
- Jones, Schonlau, Welch (1998). Efficient Global Optimization of Expensive Black-Box Functions. J. Global Optim. 13, 455 - the closed-form Expected Improvement acquisition.
- Genheden & Ryde (2015). The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities. Expert Opin. Drug Discov. 10, 449 - the MMGBSA oracle.
- Trott & Olson (2010). AutoDock Vina. J. Comput. Chem. 31, 455 - the docking score, the lower-fidelity tier.
Run it on your own data
Open it inside GDBS to save runs to Sandbox, attach results to a Worklog case, or share through a Gate client portal - all in the browser, nothing uploaded to anyone’s cloud.