Dataset Summary
DeepChem Toxicity Scores
| Registry ID | Compound Name | FDA Approved % | FDA Rejected % | Non-Toxic % | Clinical Toxicity % | Overall Risk |
|---|---|---|---|---|---|---|
| SPO-26493 | [4-[(2-chlorobenzoyl)diazenyl]anilino] (E)-3-phenylprop-2-enoate | High Risk | ||||
| SPO-30755 | 2-[2-(3-hydroxy-4-methoxyphenyl)ethyl]isoindole-1,3-dione | High Risk | ||||
| SPO-22954 | 2-[(5-carbamimidamido-1-hydroxypentylidene)amino]-3-phenylpropanoic acid | High Risk | ||||
| SPO-25901 | N-naphthalen-1-yl-N'-propylcarbamimidothioic acid | High Risk | ||||
| SPO-30413 | 4-N,5-N-didodecyl-1H-imidazole-4,5-dicarboximidic acid | Moderate Risk | ||||
| SPO-29372 | 1-(2-hydroxy-3,4,6-trimethoxyphenyl)ethanone | High Risk | ||||
| SPO-23743 | N-(N-phenyl-C-sulfanylcarbonimidoyl)-2-(6-piperidin-1-ylpyrimidin-4-yl)sulfanylethanehydrazonic acid | Lower Risk | ||||
| SPO-24630 | 9H-purine-6-thiol (6-Mercaptopurine) ⚠ GHS Danger | Moderate Risk | ||||
| SPO-27287 | 9-hydroxy-1,3-dimethylpyrimido[4,5-g]pteridine-2,4-dione | Moderate Risk |
Compound-Level Toxicity Profiles
Toxicity Risk Matrix
Favorable Moderate concern High concern N/A
Key Findings & Interpretation
✅ Best DeepChem Profiles
SPO-23743 — Strongest profile in the set: 62.2% FDA approved, 64.1% non-toxic, only 35.9% clinical toxicity. The piperidinyl-pyrimidinyl scaffold appears structurally favorable to DeepChem's model.
SPO-24630 (6-Mercaptopurine) — Highest FDA approval score at 78.6%, consistent with its status as an established clinical drug. Moderate clinical tox (59.4%) reflects known cytotoxicity profile.
SPO-27287 — Borderline approval score of 54.9% with moderate clinical tox, placing it in a "watch" category.
⚠️ High-Concern Compounds
SPO-30755 — Highest clinical toxicity score in the set (95.4%) and lowest FDA approval (3.5%). Despite favorable physicochemical properties, the isoindolone–phthalimide scaffold is flagged as high risk by DeepChem.
SPO-29372 — 92.5% FDA rejected, 87.1% clinical tox. Trimethoxyphenyl acetophenones are a class with known biological activity concerns.
SPO-25901 — 87.7% clinical tox despite simple structure; thiourea-naphthalene may be driving model prediction.
📌 Interpretation Notes
DeepChem's FDA approval/rejection classifier is trained on structural features of clinical drug candidates — a high "rejected" score reflects structural dissimilarity to approved drugs, not necessarily direct toxicity.
The clinical toxicity model predicts adverse clinical effects based on molecular features. These scores should be used alongside wet-lab assays (hERG, AMES, cytotoxicity) for full risk assessment.
SPO-24630 GHS Danger classification (H301, H340, H360, H370, H372) from the Sapio registry confirms experimentally observed toxicity, which partially aligns with the 59.4% predicted clinical tox.
🔬 Recommended Next Steps
• Progress SPO-23743 and SPO-27287 to wet-lab confirmation assays given favorable DeepChem profiles
• Flag SPO-30755, SPO-29372, SPO-25901, SPO-22954 for structural review — all score >85% clinical toxicity
• For SPO-24630: model-predicted moderate risk is consistent with its known clinical use — treat as reference compound
• Run DeepChem Solubility Predictions as a complementary tool for the set