技能 pytdc
💊
藥物發現研究人員需要標準化的資料集來訓練機器學習模型。PyTDC 提供經過策展的 ADME、毒性和藥物-標靶互動資料集，並附有適當的訓練-測試分割和評估預言機。
支援: Claude Codex Code(CC)
1
下載技能 ZIP
2
在 Claude 中上傳
前往 設定 → 功能 → 技能 → 上傳技能
3
開啟並開始使用
測試它
正在使用「pytdc」。 Load the AMES toxicity dataset and show me the data format
預期結果:
- Dataset loaded with 7,255 compounds for mutagenicity prediction
- Columns include Drug_ID, Drug (SMILES), and Y (binary toxicity label)
- Scaffold split applied: 5,078 train, 725 validation, 1,452 test molecules
正在使用「pytdc」。 Evaluate this molecule with the GSK3B oracle: CC(C)Cc1ccc(cc1)C(C)C(O)=O
預期結果:
- GSK3B binding score: 0.0234 (low predicted affinity)
- This SMILES represents ibuprofen, not expected to inhibit GSK3B
- Scores range from 0 to 1, with higher values indicating stronger predicted binding
安全審計
安全v4 • 1/17/2026
This skill provides documentation and templates for PyTDC, a legitimate drug discovery dataset library. All 427 static findings are false positives caused by markdown code blocks containing Python examples (detected as shell backticks), scientific terminology (DRD2, GSK3B detected as C2 keywords), and molecular/cryptographic naming overlaps. No actual security risks present.
9
已掃描檔案
3,184
分析行數
2
發現項
4
審計總數
風險因素
⚙️ 外部命令 (339)
EVALUATION_OUTPUT.json:24 references/datasets.md:10 references/datasets.md:11 references/datasets.md:12 references/datasets.md:13 references/datasets.md:14 references/datasets.md:15 references/datasets.md:16 references/datasets.md:19 references/datasets.md:20 references/datasets.md:21 references/datasets.md:24 references/datasets.md:25 references/datasets.md:26 references/datasets.md:27 references/datasets.md:28 references/datasets.md:29 references/datasets.md:30 references/datasets.md:31 references/datasets.md:34 references/datasets.md:35 references/datasets.md:36 references/datasets.md:39 references/datasets.md:40 references/datasets.md:41 references/datasets.md:46 references/datasets.md:47 references/datasets.md:48 references/datasets.md:49 references/datasets.md:50 references/datasets.md:51 references/datasets.md:54 references/datasets.md:55 references/datasets.md:56 references/datasets.md:57 references/datasets.md:58 references/datasets.md:59 references/datasets.md:62 references/datasets.md:63 references/datasets.md:64 references/datasets.md:65 references/datasets.md:66 references/datasets.md:67 references/datasets.md:68 references/datasets.md:69 references/datasets.md:70 references/datasets.md:71 references/datasets.md:72 references/datasets.md:73 references/datasets.md:78 references/datasets.md:79 references/datasets.md:80 references/datasets.md:83 references/datasets.md:84 references/datasets.md:85 references/datasets.md:86 references/datasets.md:87 references/datasets.md:91 references/datasets.md:92 references/datasets.md:93 references/datasets.md:97 references/datasets.md:98 references/datasets.md:102 references/datasets.md:103 references/datasets.md:107 references/datasets.md:108 references/datasets.md:112 references/datasets.md:119 references/datasets.md:120 references/datasets.md:121 references/datasets.md:124 references/datasets.md:125 references/datasets.md:128 references/datasets.md:129 references/datasets.md:133 references/datasets.md:134 references/datasets.md:138 references/datasets.md:139 references/datasets.md:143 references/datasets.md:144 references/datasets.md:148 references/datasets.md:149 references/datasets.md:153 references/datasets.md:154 references/datasets.md:155 references/datasets.md:159 references/datasets.md:160 references/datasets.md:164 references/datasets.md:168 references/datasets.md:172 references/datasets.md:176 references/datasets.md:182 references/datasets.md:183 references/datasets.md:184 references/datasets.md:185 references/datasets.md:189 references/datasets.md:190 references/datasets.md:194 references/datasets.md:195 references/datasets.md:201-209 references/datasets.md:209-215 references/datasets.md:215-224 references/datasets.md:224-230 references/datasets.md:230-239 references/oracles.md:16-27 references/oracles.md:27-40 references/oracles.md:40-43 references/oracles.md:43-49 references/oracles.md:49-52 references/oracles.md:52-58 references/oracles.md:58-61 references/oracles.md:61-67 references/oracles.md:67-70 references/oracles.md:70-76 references/oracles.md:76-79 references/oracles.md:79-85 references/oracles.md:85-88 references/oracles.md:88-93 references/oracles.md:93-96 references/oracles.md:96-101 references/oracles.md:101-104 references/oracles.md:104-109 references/oracles.md:109-112 references/oracles.md:112-117 references/oracles.md:117-120 references/oracles.md:120-131 references/oracles.md:131-134 references/oracles.md:134-140 references/oracles.md:140-143 references/oracles.md:143-151 references/oracles.md:151-154 references/oracles.md:154-160 references/oracles.md:160-163 references/oracles.md:163-169 references/oracles.md:169-172 references/oracles.md:172-181 references/oracles.md:181-184 references/oracles.md:184-189 references/oracles.md:189-192 references/oracles.md:192-197 references/oracles.md:197-200 references/oracles.md:200-205 references/oracles.md:205-208 references/oracles.md:208-213 references/oracles.md:213-216 references/oracles.md:216-221 references/oracles.md:221-224 references/oracles.md:224-231 references/oracles.md:231-234 references/oracles.md:234-240 references/oracles.md:240-243 references/oracles.md:243-249 references/oracles.md:249-252 references/oracles.md:252-261 references/oracles.md:261-280 references/oracles.md:280-295 references/oracles.md:295-300 references/oracles.md:300-305 references/oracles.md:305-327 references/oracles.md:327-330 references/oracles.md:330-350 references/oracles.md:350-354 references/oracles.md:354-366 references/oracles.md:366-380 references/oracles.md:380-394 references/utilities.md:19-34 references/utilities.md:34-41 references/utilities.md:41-43 references/utilities.md:43-57 references/utilities.md:57-59 references/utilities.md:59-76 references/utilities.md:76-80 references/utilities.md:80-85 references/utilities.md:85-87 references/utilities.md:87-92 references/utilities.md:92-94 references/utilities.md:94-101 references/utilities.md:101-103 references/utilities.md:103-112 references/utilities.md:112-118 references/utilities.md:118-124 references/utilities.md:124-126 references/utilities.md:126-136 references/utilities.md:136-144 references/utilities.md:144-151 references/utilities.md:151-154 references/utilities.md:154-166 references/utilities.md:166-169 references/utilities.md:169-181 references/utilities.md:181-184 references/utilities.md:184-195 references/utilities.md:195-198 references/utilities.md:198-209 references/utilities.md:209-212 references/utilities.md:212-219 references/utilities.md:219-222 references/utilities.md:222-231 references/utilities.md:231-234 references/utilities.md:234-243 references/utilities.md:243-246 references/utilities.md:246-255 references/utilities.md:255-258 references/utilities.md:258-267 references/utilities.md:267-270 references/utilities.md:270-282 references/utilities.md:282-285 references/utilities.md:285-295 references/utilities.md:295-298 references/utilities.md:298-300 references/utilities.md:300 references/utilities.md:300 references/utilities.md:300 references/utilities.md:300-306 references/utilities.md:306-321 references/utilities.md:321-331 references/utilities.md:331-345 references/utilities.md:345-354 references/utilities.md:354-357 references/utilities.md:357-363 references/utilities.md:363-379 references/utilities.md:379-382 references/utilities.md:382-383 references/utilities.md:383-384 references/utilities.md:384-385 references/utilities.md:385-386 references/utilities.md:386-387 references/utilities.md:387-391 references/utilities.md:391-397 references/utilities.md:397-405 references/utilities.md:405-412 references/utilities.md:412-418 references/utilities.md:418-426 references/utilities.md:426-437 references/utilities.md:437-442 references/utilities.md:442-448 references/utilities.md:448-456 references/utilities.md:456-462 references/utilities.md:462-470 references/utilities.md:470-476 references/utilities.md:476-486 references/utilities.md:486-498 references/utilities.md:498-503 references/utilities.md:503-506 references/utilities.md:506-511 references/utilities.md:511-514 references/utilities.md:514-520 references/utilities.md:520-526 references/utilities.md:526-535 references/utilities.md:535-541 references/utilities.md:541-548 references/utilities.md:548-552 references/utilities.md:552-564 references/utilities.md:564-568 references/utilities.md:568-580 references/utilities.md:580-586 references/utilities.md:586-615 references/utilities.md:615-619 references/utilities.md:619-637 references/utilities.md:637-641 references/utilities.md:641-661 SKILL.md:30-32 SKILL.md:32-36 SKILL.md:36-38 SKILL.md:38-49 SKILL.md:49-54 SKILL.md:54-57 SKILL.md:57 SKILL.md:57 SKILL.md:57 SKILL.md:57-58 SKILL.md:58-59 SKILL.md:59-63 SKILL.md:63-68 SKILL.md:68-80 SKILL.md:80-84 SKILL.md:84-99 SKILL.md:99-103 SKILL.md:103-116 SKILL.md:116-119 SKILL.md:119-125 SKILL.md:125-128 SKILL.md:128-140 SKILL.md:140 SKILL.md:140-141 SKILL.md:141 SKILL.md:141-142 SKILL.md:142-154 SKILL.md:154-158 SKILL.md:158-172 SKILL.md:172-176 SKILL.md:176-184 SKILL.md:184-187 SKILL.md:187-208 SKILL.md:208-212 SKILL.md:212-216 SKILL.md:216-220 SKILL.md:220-222 SKILL.md:222-228 SKILL.md:228-232 SKILL.md:232-240 SKILL.md:240-243 SKILL.md:243-245 SKILL.md:245 SKILL.md:245-253 SKILL.md:253-268 SKILL.md:268-280 SKILL.md:280-290 SKILL.md:290-300 SKILL.md:300-303 SKILL.md:303-304 SKILL.md:304-305 SKILL.md:305 SKILL.md:305 SKILL.md:305-306 SKILL.md:306-312 SKILL.md:312-322 SKILL.md:322-330 SKILL.md:330-336 SKILL.md:336-347 SKILL.md:347-353 SKILL.md:353-363 SKILL.md:363-365 SKILL.md:365-371 SKILL.md:371-379 SKILL.md:379-383 SKILL.md:383-390 SKILL.md:390-394 SKILL.md:394-402 SKILL.md:402-408 SKILL.md:408-410 SKILL.md:410-426 SKILL.md:426-430 SKILL.md:430-434 SKILL.md:434-442 SKILL.md:442-443 SKILL.md:443-444 SKILL.md:444-448 SKILL.md:448-449 SKILL.md:449-450
🌐 網路存取 (15)
EVALUATION_OUTPUT.json:42 EVALUATION_OUTPUT.json:42 EVALUATION_OUTPUT.json:42 references/datasets.md:246 references/oracles.md:374 references/oracles.md:398 references/utilities.md:681 references/utilities.md:682 references/utilities.md:683 references/utilities.md:684 scripts/benchmark_evaluation.py:268 skill-report.json:6 SKILL.md:454 SKILL.md:455 SKILL.md:456
審計者: claude 查看審計歷史 →
品質評分
68
架構
100
可維護性
87
內容
19
社群
100
安全
83
規範符合性
你能建構什麼
訓練 ADME 預測模型
使用支架分割載入 Caco-2 滲透性資料集，訓練分子屬性預測器，並使用標準指標進行評估。
評估毒性預測模型
存取 hERG、AMES 和 DILI 毒性資料集，配合基準協議來驗證安全性預測模型。
生成新型藥物候選物
使用分子預言機如 GSK3B 和 DRD2 來引導生成模型，產生具有所需生物活性的化合物。
試試這些提示
載入 ADME 資料集
幫助我從 TDC 載入 Caco2_Wang 資料集，並使用支架分割來訓練腸道滲透性預測模型。
執行基準評估
展示如何使用 TDC 基準群組和所需的 5 種子協議來評估我的 ADME 模型。
使用分子預言機
我想使用 TDC 預言機來評估生成的 SMILES 字串，針對 QED、SA 和 GSK3B 屬性。展示我工作流程。
藥物-標靶互動建模
載入 BindingDB_Kd 資料集，使用冷藥物分割來確保我的模型能推廣到未見過的藥物化合物。
最佳實務
- 使用支架分割而非隨機分割，以在新型化學支架上實現更真實的模型評估
- 使用所有 5 個必要種子執行基準評估，以報告平均和標準差效能
- 結合多個預言機與加權評分以實現多目標分子優化
避免
- 避免對生產環境 ADME 模型使用隨機分割，因為它們會高估相似分子的效能
- 不要報告單一種子基準結果，因為它們可能無法反映真正的模型變異數
- 避免使用預言機作為訓練的真實標籤，因為它們本身是預測模型
常見問題
PyTDC 中有哪些可用的資料集？
PyTDC 包含 60 多個涵蓋 ADME、毒性、藥物-標靶互動、藥物-藥物互動和分子生成任務的資料集，專為治療性機器學習設計。
什麼是支架分割，為什麼要使用它？
支架分割根據化學支架對分子進行分組，使測試分子具有與訓練分子不同的核心結構，模擬真實世界的推廣能力。
分子預言機如何運作？
預言機是預訓練模型，可對 SMILES 字串進行評分，針對藥物相似性 (QED)、合成可及性 (SA) 或標靶結合 (GSK3B、DRD2) 等屬性。
什麼是基準評估的 5 種子協議？
TDC 基準評估需要使用 5 個不同的隨機種子來計算平均值和標準差，確保穩健的效能比較。
我可以使用 PyTDC 搭配 PyTorch Geometric 或 DGL 嗎？
是的，TDC 提供 MolConvert 工具，可以將 SMILES 轉換為 PyG 圖形、DGL 圖形或其他分子表示，如 ECFP 指紋。
什麼是 DTI 預測的冷藥物分割？
冷藥物分割確保測試集中的藥物永遠不會出現在訓練集中，測量模型對全新藥物化合物預測結合的能力。