Update CYP challenge method report
Browse files
README.md
CHANGED
|
@@ -6,44 +6,73 @@ tags:
|
|
| 6 |
---
|
| 7 |
# yw — OpenADMET CYP direct-inhibition method report
|
| 8 |
|
| 9 |
-
Updated: 2026-
|
| 10 |
|
| 11 |
## Submission
|
| 12 |
|
| 13 |
-
This report describes the
|
| 14 |
-
|
| 15 |
|
| 16 |
-
Prediction file SHA-256: `
|
| 17 |
-
Model revision: `06e1089342f4ec8297830b103935684e73668812`.
|
| 18 |
|
| 19 |
-
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
-
|
| 22 |
|
| 23 |
-
|
| 24 |
|
| 25 |
-
|
| 26 |
-
-
|
| 27 |
-
|
| 28 |
-
|
|
|
|
|
|
|
| 29 |
|
| 30 |
-
|
| 31 |
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
as distinct assay observations. CYP2D6 direct measurements are selected
|
| 35 |
-
from screening hits, limiting what random-fold direct validation establishes.
|
| 36 |
|
| 37 |
-
##
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
|
| 39 |
-
|
| 40 |
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
-
|
| 49 |
-
|
|
|
|
|
|
|
|
|
| 6 |
---
|
| 7 |
# yw — OpenADMET CYP direct-inhibition method report
|
| 8 |
|
| 9 |
+
Updated: 2026-11-03T16:30:42.279580+00:00
|
| 10 |
|
| 11 |
## Submission
|
| 12 |
|
| 13 |
+
This report describes the final public challenge submission under the anonymous leaderboard
|
| 14 |
+
alias **yw**. It uses no proprietary training data and no external assay data. No code is published.
|
| 15 |
|
| 16 |
+
Prediction file SHA-256: `80695a19012518a3fbfd5ec3086dd112701b307b27abacba41b9472ced05ed0f`.
|
|
|
|
| 17 |
|
| 18 |
+
The file combines, per enzyme, prediction columns from earlier model runs (no joint refit and no new
|
| 19 |
+
training). Each enzyme's column was chosen by comparing the public live-leaderboard ST-RAE of earlier
|
| 20 |
+
submissions that contained it. The live leaderboard covers only part of the blinded series, so these
|
| 21 |
+
choices are tuned to that subset.
|
| 22 |
|
| 23 |
+
## Live leaderboard result
|
| 24 |
|
| 25 |
+
Overall MA-ST-RAE of this file: expected 0.7404 (mean of the four per-enzyme scores; the board value will be added when scored).
|
| 26 |
|
| 27 |
+
| Enzyme | Source run | Live ST-RAE of the column |
|
| 28 |
+
| --- | --- | ---: |
|
| 29 |
+
| CYP1A2 | `20261031T032022-cyp1a2-nested-support-conditioned-family-gate` | 0.772 |
|
| 30 |
+
| CYP2C9 | `20261029T175417-cyp1a2-cyp3a4-fixed-descriptor-graph-ensemble` | 0.5217 |
|
| 31 |
+
| CYP2D6 | `20261017T004159-cyp2d6-assay-tagged-nonhit-calibration` | 1.172 |
|
| 32 |
+
| CYP3A4 | `20261029T175417-cyp1a2-cyp3a4-fixed-descriptor-graph-ensemble` | 0.4958 |
|
| 33 |
|
| 34 |
+
## Data and method of each source run
|
| 35 |
|
| 36 |
+
Each description is quoted from that run's own record. Only the enzymes listed for a run are taken
|
| 37 |
+
from it, so statements about its other enzymes do not apply to this file.
|
|
|
|
|
|
|
| 38 |
|
| 39 |
+
### `20261031T032022-cyp1a2-nested-support-conditioned-family-gate` (used for CYP1A2)
|
| 40 |
+
|
| 41 |
+
Model revision: `c232bdc94f87c336f3253b2e51c52bae87474d32`.
|
| 42 |
+
|
| 43 |
+
Method: CYP1A2 p=L+clip(0.5+Z beta,0,1)*(J-L), fixed eight-column support gate. Deterministic interval-error LP plus .01 L1 beta penalty, training weights constrained [0,1]. Full real-support gate is sole candidate; no control selection. Three other endpoints exactly F.
|
| 44 |
+
|
| 45 |
+
Features: intercept; m_z; d_z; abs_d_z; similarity_z; scaffold; m_z*d_z; similarity_z*d_z
|
| 46 |
+
|
| 47 |
+
Data sources: Challenge direct confidence intervals; Archived L tree and J descriptor-graph nested predictions; separately tagged TDI-condition screening supervision only in base models
|
| 48 |
+
|
| 49 |
+
Validation: Controller canonical random/scaffold five-fold, testlike holdout and archived Butina cluster five-fold. Gate training uses only each population inner OOF archive and permitted fitting labels; blinded fit uses full inner OOF. All 85 inner fits and 17 refits audited before gate training; no global outer OOF input.
|
| 50 |
+
|
| 51 |
+
### `20261029T175417-cyp1a2-cyp3a4-fixed-descriptor-graph-ensemble` (used for CYP2C9, CYP3A4)
|
| 52 |
+
|
| 53 |
+
Model revision: `911812692f61783085b34085e0fb498683c28dcc`.
|
| 54 |
|
| 55 |
+
Method: Fixed (L + GJ)/2 for CYP1A2 and CYP3A4; C CYP2C9/CYP2D6 copied exactly. GJ is the archived equal three-seed epoch-100 graph. No training, calibration, learned weights, checkpoint selection, or control substitution.
|
| 56 |
|
| 57 |
+
Features: Chemprop 2.3.1 depth-3 256-dimensional BondMessagePassing, default atom/bond features, fixed half-mean/half-componentwise-max pooling; X_d concatenation to 268-input two-layer eight-output RegressionFFN, hidden256, dropout0.1, no batch normalization; ['MolWt', 'MolLogP', 'TPSA', 'MolMR', 'NumHDonors', 'NumHAcceptors', 'NumRotatableBonds', 'RingCount', 'NumAromaticRings', 'FractionCSP3', 'HeavyAtomCount', 'NumHeteroatoms']; Historical LightGBM: chiral Morgan counts, RDKit descriptors, nested predicted TDI-condition screening responses
|
| 58 |
+
|
| 59 |
+
Data sources: Challenge direct-inhibition intervals; Separate TDI-condition single-concentration screen tasks; Hash-pinned completed L, C, and J archives; no new assay data
|
| 60 |
+
|
| 61 |
+
Validation: Unchanged canonical random/scaffold five-fold and testlike holdout; C fixed Butina cluster diagnostic. Every graph fit/held index and fitting-only descriptor/target normalization verified; all 153 saved inference arrays exactly reproduce seed-average exports. Historical stacking control flow and saved cluster screen memberships audited. TDI/Emax labels were not used in these components. All four predictions finite for every held row. No held assay enters a fitted component.
|
| 62 |
+
|
| 63 |
+
### `20261017T004159-cyp2d6-assay-tagged-nonhit-calibration` (used for CYP2D6)
|
| 64 |
+
|
| 65 |
+
Model revision: `06e1089342f4ec8297830b103935684e73668812`.
|
| 66 |
+
|
| 67 |
+
Method: Interval-bound LightGBM champion with a CYP2D6 assay-tagged screen nonhit probability correction
|
| 68 |
+
|
| 69 |
+
Features: Morgan radius-2 count fingerprints with chirality, RDKit descriptors, a CYP2D6 screen nonhit classifier and separate hit/nonhit L1 response models
|
| 70 |
+
|
| 71 |
+
Data sources: OpenADMET direct-inhibition training data (champion); OpenADMET single-concentration log2 fold change under TDI condition; Local screen-multitask-features champion predictions
|
| 72 |
+
|
| 73 |
+
Validation: Controller molecule-grouped random, scaffold, test-like direct splits and five-fold screen splits; all screen data from held-out canonical molecules excluded
|
| 74 |
|
| 75 |
+
Compounds were separated before fitting any model component or generating out-of-fold assay
|
| 76 |
+
features. Direct, TDI, and screening readouts were treated as distinct assay observations. CYP2D6
|
| 77 |
+
direct measurements are selected from screening hits, limiting what random-fold direct validation
|
| 78 |
+
establishes.
|