AlphaFold DB 提供超過 2 億個 AI 預測的蛋白質結構。此技能協助研究人員檢索結構檔案、分析信賴度分數,並將預測整合到藥物發現和結構生物學的運算流程中。
下载技能 ZIP
在 Claude 中上传
前往 设置 → 功能 → 技能 → 上传技能
开启并开始使用
测试它
正在使用“alphafold-database”。 Get the AlphaFold structure for protein ABL1 (P00520) and analyze its confidence scores
预期结果:
- Protein: ABL1 (Tyrosine-protein kinase ABL1)
- AlphaFold ID: AF-P00520-F1
- Sequence length: 1130 residues
- Average pLDDT: 85 (high confidence)
- High confidence regions: Kinase domain residues show very high confidence (pLDDT > 90)
- Low confidence regions: N-terminal disordered regions show lower confidence (pLDDT < 50)
- Files available: model_v4.cif, confidence_v4.json, predicted_aligned_error_v4.json
正在使用“alphafold-database”。 Find high-confidence human proteins in AlphaFold DB
预期结果:
- Query results from BigQuery metadata table:
- Found proteins with fractionPlddtVeryHigh > 0.8 in human proteome
- Top results include well-structured cytosolic proteins
- Average sequence length varies from 300 to 2500 residues
- All predictions meet AlphaFold quality thresholds
安全审计
安全Documentation-only skill for accessing the AlphaFold Protein Structure Database. Contains markdown documentation with Python code examples for legitimate scientific APIs. All network endpoints are public scientific databases (alphafold.ebi.ac.uk, uniprot.org, Google Cloud). No executable code, no sensitive data access, no malicious patterns found after human review.
风险因素
🌐 网络访问 (1)
⚙️ 外部命令 (1)
质量评分
你能构建什么
獲取蛋白質模型
下載沒有實驗資料的蛋白質預測結構,用於比較分析。
分析結合位點
檢索目標蛋白質結構並評估虛擬篩選研究的信賴度指標。
建立分析流程
將 AlphaFold API 呼叫整合到大規模蛋白質體處理的自動化工作流程中。
试试这些提示
Get the AlphaFold structure for UniProt ID P00520. Download the mmCIF file and summarize the protein.
Compare pLDDT confidence scores between proteins P12931 and P04637. Which has more high-confidence regions?
Find all human proteins in AlphaFold DB with very high confidence (pLDDT greater than 90) using BigQuery. List the top 10 by sequence length.
Download the PAE matrix for protein P00520 and explain the domain arrangement confidence based on the error values.
最佳实践
- 在本地快取已下載的結構檔案,避免重複的 API 請求
- 使用 Google Cloud Storage 進行大量蛋白質體下載,而非個別的 REST API 呼叫
- 在進行下游結構分析前驗證預測信賴度(pLDDT > 70)
避免
- 使用 REST API 進行大量下載(效率低,可能達到速率限制)
- 假設所有預測都同樣可靠(務必檢查信賴度分數)
- 在分析多結構域蛋白質排列時忽略 PAE 值