yw — OpenADMET CYP direct-inhibition method report
Updated: 2026-10-16T04:22:12.588223+00:00
Submission
This report describes the latest best-scored public challenge submission under the anonymous leaderboard alias yw. It uses no proprietary training data.
Prediction file SHA-256: 89fe94524bbfa4650c3e365d681709c9b11a9f5070688fa2b28862450a95d51a.
Model revision: c09db1237c698b877a9740a80fa82f3a610be423.
Data and method
Interval-bound LightGBM champion with a conservative CYP2D6 screen-response selection correction
Features: Morgan radius-2 count fingerprints with chirality and RDKit descriptors
Data sources:
- OpenADMET direct-inhibition training data (champion)
- OpenADMET single-concentration log2 fold change under TDI condition
Validation: Controller molecule-grouped random, scaffold, test-like direct splits and five-fold screen splits; all screen data from held-out canonical molecules excluded
Compounds were separated before fitting any model component or generating out-of-fold assay features. Direct, TDI, and screening readouts were treated as distinct assay observations. CYP2D6 direct measurements are selected from screening hits, limiting what random-fold direct validation establishes.
Live leaderboard result
Overall MA-ST-RAE: 0.8326.
| Enzyme | ST-RAE |
|---|---|
| CYP1A2 | 0.9466 |
| CYP2C9 | 0.5811 |
| CYP2D6 | 1.1904 |
| CYP3A4 | 0.6124 |
The live leaderboard uses a blinded subset. The report will be updated when the final prediction is selected.