[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6DNE1WF71SC9MKCTFBJBVPW":3,"skill-same:01M6DNE1WF71SC9MKCTFBJBVPW":59},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":31,"linkOnly":32,"markdownContent":33,"markdownBody":34,"frontMatter":35,"refName":40,"commitSha":41,"skillDirectory":5,"forksCount":14,"repositoryHtmlUrl":42,"repositoryHomepageUrl":6,"ownerHtmlUrl":43,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":44,"install":45,"resources":58},"01M6DNE1WF71SC9MKCTFBJBVPW","boltz",null,"Structure prediction using Boltz-1\u002FBoltz-2, an open biomolecular structure predictor. Use this skill when: (1) Predicting protein complex structures, (2) Validating designed binders, (3) Need open-source alternative to AF2, (4) Predicting protein-ligand complexes, (5) Using local GPU resources.\nFor QC thresholds, use protein-qc. For AlphaFold2 prediction, use alphafold. For Chai prediction, use chai.","huamu668\u002Fclawhub-openclaw-medical","huamu668","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F202473531?v=4","01M6DK1V7D7VP9PJ06Z1VXVK13","huamu","01M6DK1VBYG1BX5FT154AFZHAS",0,"MIT","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fboltz\u002FSKILL.md",[],[5,19,20,21,22,23,24,25,26,27,28,29,30],"design","validation","gpu","Tools","modal","open","use","structure","local","structure-prediction","alphafold","open-source","2026-11-10 10:12:30",false,"---\nname: boltz\ndescription: >\n  Structure prediction using Boltz-1\u002FBoltz-2, an open biomolecular structure predictor.\n  Use this skill when: (1) Predicting protein complex structures,\n  (2) Validating designed binders,\n  (3) Need open-source alternative to AF2,\n  (4) Predicting protein-ligand complexes,\n  (5) Using local GPU resources.\n\n  For QC thresholds, use protein-qc.\n  For AlphaFold2 prediction, use alphafold.\n  For Chai prediction, use chai.\nlicense: MIT\ncategory: design-tools\ntags: [structure-prediction, validation, open-source]\nbiomodals_script: modal_boltz.py\n---\n\n# Boltz Structure Prediction\n\n## Prerequisites\n\n| Requirement | Minimum | Recommended |\n|-------------|---------|-------------|\n| Python | 3.10+ | 3.11 |\n| CUDA | 12.0+ | 12.1+ |\n| GPU VRAM | 24GB | 48GB (L40S) |\n| RAM | 32GB | 64GB |\n\n## How to run\n\n> **First time?** See [Installation Guide](..\u002F..\u002Fdocs\u002Finstallation.md) to set up Modal and biomodals.\n\n### Option 1: Modal\n```bash\ncd biomodals\nmodal run modal_boltz.py \\\n  --input-faa complex.fasta \\\n  --out-dir predictions\u002F\n```\n\n**GPU**: L40S (48GB) | **Timeout**: 1800s default\n\n### Option 2: Local installation\n```bash\npip install boltz\n\nboltz predict \\\n  --fasta complex.fasta \\\n  --output predictions\u002F\n```\n\n## Key parameters\n\n| Parameter | Default | Range | Description |\n|-----------|---------|-------|-------------|\n| `--recycling_steps` | 3 | 1-10 | Recycling iterations |\n| `--sampling_steps` | 200 | 50-500 | Diffusion steps |\n| `--use_msa_server` | true | bool | Use MSA server |\n\n## FASTA Format\n\n```\n>protein_A\nMKTAYIAKQRQISFVK...\n>protein_B\nMVLSPADKTNVKAAWG...\n```\n\n## Output format\n\n```\npredictions\u002F\n├── model_0.cif       # Best model (CIF format)\n├── confidence.json   # pLDDT, pTM, ipTM\n└── pae.npy          # PAE matrix\n```\n\n**Note**: Boltz outputs CIF format. Convert to PDB if needed:\n```python\nfrom Bio.PDB import MMCIFParser, PDBIO\nparser = MMCIFParser()\nstructure = parser.get_structure(\"model\", \"model_0.cif\")\nio = PDBIO()\nio.set_structure(structure)\nio.save(\"model_0.pdb\")\n```\n\n## Comparison\n\n| Feature | Boltz-1 | Boltz-2 | AF2-Multimer |\n|---------|---------|---------|--------------|\n| MSA-free mode | Yes | Yes | No |\n| Diffusion | Yes | Yes | No |\n| Speed | Fast | Faster | Slower |\n| Open source | Yes | Yes | Yes |\n\n## Sample output\n\n### Successful run\n```\n$ boltz predict --fasta complex.fasta --output predictions\u002F\n[INFO] Loading Boltz-1 weights...\n[INFO] Predicting structure...\n[INFO] Saved model to predictions\u002Fmodel_0.cif\n\npredictions\u002Fconfidence.json:\n{\n  \"ptm\": 0.78,\n  \"iptm\": 0.65,\n  \"plddt\": 0.81\n}\n```\n\n**What good output looks like:**\n- pTM: > 0.7 (confident global structure)\n- ipTM: > 0.5 (confident interface)\n- pLDDT: > 0.7 (confident per-residue)\n- CIF file: ~100-500 KB for typical complex\n\n## Decision tree\n\n```\nShould I use Boltz?\n│\n├─ What are you predicting?\n│  ├─ Protein-protein complex → Boltz ✓ or Chai or ColabFold\n│  ├─ Protein + ligand → Boltz ✓ or Chai\n│  └─ Single protein → Use ESMFold (faster)\n│\n├─ Need MSA?\n│  ├─ No \u002F want speed → Boltz ✓\n│  └─ Yes \u002F maximum accuracy → ColabFold\n│\n└─ Why Boltz over Chai?\n   ├─ Open weights preference → Boltz ✓\n   ├─ Boltz-2 speed → Boltz ✓\n   └─ DNA\u002FRNA support → Consider Chai\n```\n\n## Typical performance\n\n| Campaign Size | Time (L40S) | Cost (Modal) | Notes |\n|---------------|-------------|--------------|-------|\n| 100 complexes | 30-45 min | ~$8 | Standard validation |\n| 500 complexes | 2-3h | ~$35 | Large campaign |\n| 1000 complexes | 4-6h | ~$70 | Comprehensive |\n\n**Per-complex**: ~15-30s for typical binder-target complex.\n\n---\n\n## Verify\n\n```bash\nfind predictions -name \"*.cif\" | wc -l  # Should match input count\n```\n\n---\n\n## Troubleshooting\n\n**Low confidence**: Increase recycling_steps\n**OOM errors**: Use MSA-free mode or A100-80GB\n**Slow prediction**: Reduce sampling_steps\n\n### Error interpretation\n\n| Error | Cause | Fix |\n|-------|-------|-----|\n| `RuntimeError: CUDA out of memory` | Complex too large | Use `--use_msa_server false` or larger GPU |\n| `KeyError: 'iptm'` | Single chain only | Ensure FASTA has 2+ chains |\n| `FileNotFoundError: weights` | Missing model | Run `boltz download` first |\n| `ValueError: invalid residue` | Non-standard AA | Check for modified residues in sequence |\n\n### Boltz-1 vs Boltz-2\n\n| Aspect | Boltz-1 | Boltz-2 |\n|--------|---------|---------|\n| Speed | Fast | ~2x faster |\n| Accuracy | Good | Improved |\n| Ligands | Basic | Better support |\n| Release | 2024 | Late 2024 |\n\n---\n\n**Next**: `protein-qc` for filtering and ranking.\n","# Boltz Structure Prediction\n\n## Prerequisites\n\n| Requirement | Minimum | Recommended |\n|-------------|---------|-------------|\n| Python | 3.10+ | 3.11 |\n| CUDA | 12.0+ | 12.1+ |\n| GPU VRAM | 24GB | 48GB (L40S) |\n| RAM | 32GB | 64GB |\n\n## How to run\n\n> **First time?** See [Installation Guide](..\u002F..\u002Fdocs\u002Finstallation.md) to set up Modal and biomodals.\n\n### Option 1: Modal\n```bash\ncd biomodals\nmodal run modal_boltz.py \\\n  --input-faa complex.fasta \\\n  --out-dir predictions\u002F\n```\n\n**GPU**: L40S (48GB) | **Timeout**: 1800s default\n\n### Option 2: Local installation\n```bash\npip install boltz\n\nboltz predict \\\n  --fasta complex.fasta \\\n  --output predictions\u002F\n```\n\n## Key parameters\n\n| Parameter | Default | Range | Description |\n|-----------|---------|-------|-------------|\n| `--recycling_steps` | 3 | 1-10 | Recycling iterations |\n| `--sampling_steps` | 200 | 50-500 | Diffusion steps |\n| `--use_msa_server` | true | bool | Use MSA server |\n\n## FASTA Format\n\n```\n>protein_A\nMKTAYIAKQRQISFVK...\n>protein_B\nMVLSPADKTNVKAAWG...\n```\n\n## Output format\n\n```\npredictions\u002F\n├── model_0.cif       # Best model (CIF format)\n├── confidence.json   # pLDDT, pTM, ipTM\n└── pae.npy          # PAE matrix\n```\n\n**Note**: Boltz outputs CIF format. Convert to PDB if needed:\n```python\nfrom Bio.PDB import MMCIFParser, PDBIO\nparser = MMCIFParser()\nstructure = parser.get_structure(\"model\", \"model_0.cif\")\nio = PDBIO()\nio.set_structure(structure)\nio.save(\"model_0.pdb\")\n```\n\n## Comparison\n\n| Feature | Boltz-1 | Boltz-2 | AF2-Multimer |\n|---------|---------|---------|--------------|\n| MSA-free mode | Yes | Yes | No |\n| Diffusion | Yes | Yes | No |\n| Speed | Fast | Faster | Slower |\n| Open source | Yes | Yes | Yes |\n\n## Sample output\n\n### Successful run\n```\n$ boltz predict --fasta complex.fasta --output predictions\u002F\n[INFO] Loading Boltz-1 weights...\n[INFO] Predicting structure...\n[INFO] Saved model to predictions\u002Fmodel_0.cif\n\npredictions\u002Fconfidence.json:\n{\n  \"ptm\": 0.78,\n  \"iptm\": 0.65,\n  \"plddt\": 0.81\n}\n```\n\n**What good output looks like:**\n- pTM: > 0.7 (confident global structure)\n- ipTM: > 0.5 (confident interface)\n- pLDDT: > 0.7 (confident per-residue)\n- CIF file: ~100-500 KB for typical complex\n\n## Decision tree\n\n```\nShould I use Boltz?\n│\n├─ What are you predicting?\n│  ├─ Protein-protein complex → Boltz ✓ or Chai or ColabFold\n│  ├─ Protein + ligand → Boltz ✓ or Chai\n│  └─ Single protein → Use ESMFold (faster)\n│\n├─ Need MSA?\n│  ├─ No \u002F want speed → Boltz ✓\n│  └─ Yes \u002F maximum accuracy → ColabFold\n│\n└─ Why Boltz over Chai?\n   ├─ Open weights preference → Boltz ✓\n   ├─ Boltz-2 speed → Boltz ✓\n   └─ DNA\u002FRNA support → Consider Chai\n```\n\n## Typical performance\n\n| Campaign Size | Time (L40S) | Cost (Modal) | Notes |\n|---------------|-------------|--------------|-------|\n| 100 complexes | 30-45 min | ~$8 | Standard validation |\n| 500 complexes | 2-3h | ~$35 | Large campaign |\n| 1000 complexes | 4-6h | ~$70 | Comprehensive |\n\n**Per-complex**: ~15-30s for typical binder-target complex.\n\n---\n\n## Verify\n\n```bash\nfind predictions -name \"*.cif\" | wc -l  # Should match input count\n```\n\n---\n\n## Troubleshooting\n\n**Low confidence**: Increase recycling_steps\n**OOM errors**: Use MSA-free mode or A100-80GB\n**Slow prediction**: Reduce sampling_steps\n\n### Error interpretation\n\n| Error | Cause | Fix |\n|-------|-------|-----|\n| `RuntimeError: CUDA out of memory` | Complex too large | Use `--use_msa_server false` or larger GPU |\n| `KeyError: 'iptm'` | Single chain only | Ensure FASTA has 2+ chains |\n| `FileNotFoundError: weights` | Missing model | Run `boltz download` first |\n| `ValueError: invalid residue` | Non-standard AA | Check for modified residues in sequence |\n\n### Boltz-1 vs Boltz-2\n\n| Aspect | Boltz-1 | Boltz-2 |\n|--------|---------|---------|\n| Speed | Fast | ~2x faster |\n| Accuracy | Good | Improved |\n| Ligands | Basic | Better support |\n| Release | 2024 | Late 2024 |\n\n---\n\n**Next**: `protein-qc` for filtering and ranking.\n",{"name":5,"tags":36,"license":15,"category":37,"description":38,"biomodals_script":39},[28,20,30],"design-tools","Structure prediction using Boltz-1\u002FBoltz-2, an open biomolecular structure predictor. Use this skill when: (1) Predicting protein complex structures, (2) Validating designed binders, (3) Need open-source alternative to AF2, (4) Predicting protein-ligand complexes, (5) Using local GPU resources.\nFor QC thresholds, use protein-qc. For AlphaFold2 prediction, use alphafold. For Chai prediction, use chai.\n","modal_boltz.py","main","d3998984163a67a891ee8397e35adf79b14e872d","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical","https:\u002F\u002Fgithub.com\u002Fhuamu668","2026-03-12 11:49:42",{"source":8,"skillName":5,"commands":46,"prompt":56,"zipAvailable":57},[47,50,53],{"tool":48,"command":49},"npx","npx skills add huamu668\u002Fclawhub-openclaw-medical --skill boltz",{"tool":51,"command":52},"bunx","bunx skills add huamu668\u002Fclawhub-openclaw-medical --skill boltz",{"tool":54,"command":55},"pnpm","pnpm dlx skills add huamu668\u002Fclawhub-openclaw-medical --skill boltz","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fboltz\u002FSKILL.md",true,[],[60,80,110,130,144,170],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":31},"01M6DR3JGAEX5G5FR7TK3D8NS8","zarr-python","Chunked N-D arrays for cloud storage. Compressed arrays, parallel I\u002FO, S3\u002FGCS integration, NumPy\u002FDask\u002FXarray compatible, for large-scale scientific computing pipelines.","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fzarr-python\u002FSKILL.md",[],[67,68,69,70,71,72,73,74,75,76,77,78,79],"python","zarr python","storage","scientific computing","numpy","arrays","pipelines","parallel","dask","cloud","gcs","s3","integration",{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":91,"publishedAt":31},"01M6DR101MEDE5XT2ZZ5VBEAA5","vaex","Use this skill for processing and analyzing large tabular datasets (billions of rows) that exceed available RAM. Vaex excels at out-of-core DataFrame operations, lazy evaluation, fast aggregations, efficient visualization of big data, and machine learning on large datasets. Apply when users need to work with large CSV\u002FHDF5\u002FArrow\u002FParquet files, perform fast statistics on massive datasets, create visualizations of big data, or build ML pipelines that do not fit in memory.","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fvaex\u002FSKILL.md",[86],{"categoryNo":87,"categoryName":88,"taxonomyType":89,"taxonomyTypeDesc":90},"01JBSEEDCAT000000000000013","分析与可视化","SCENARIO","使用场景",[82,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,25,108,109],"machine learning","in memory","learning","csv","visualization","create","core","ml","big data","memory","visualizations","fit","statistics","do","build","files","operations","work",{"skillNo":111,"name":112,"nameZh":6,"description":113,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":114,"categories":115,"keywords":116,"publishedAt":31},"01M6DQVMB0GBKF2W44FKXVD1P7","torch-geometric","Graph Neural Networks (PyG). Node\u002Fgraph classification, link prediction, GCN, GAT, GraphSAGE, heterogeneous graphs, molecular property prediction, for geometric deep learning.","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Ftorch-geometric\u002FSKILL.md",[],[117,94,118,119,120,121,122,123,124,125,126,127,128,129],"torch geometric","node","Neural","link","gat","molecular","graphs","deep","networks","graph","neural networks","deep learning","classification",{"skillNo":131,"name":132,"nameZh":6,"description":133,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":136,"publishedAt":31},"01M6DQH2GZPZKEQ140GRCJ9746","tiledbvcf","Efficient storage and retrieval of genomic variant data using TileDB. Scalable VCF\u002FBCF ingestion, incremental sample addition, compressed storage, parallel queries, and export capabilities for population genomics.","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Ftiledbvcf\u002FSKILL.md",[],[132,137,138,139,140,141,142,74,143,69],"data","genomics","sample","queries","ingestion","export","capabilities",{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":158,"publishedAt":31},"01M6DQD9G958EM6NHJ994D0SK1","solublempnn","Solubility-optimized protein sequence design using SolubleMPNN. Use this skill when: (1) Designing for E. coli expression, (2) Optimizing solubility of designed proteins, (3) Reducing aggregation propensity, (4) Need high-yield expression, (5) Avoiding inclusion body formation.\nFor standard design, use proteinmpnn. For ligand-aware design, use ligandmpnn.","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fsolublempnn\u002FSKILL.md",[150,153],{"categoryNo":151,"categoryName":152,"taxonomyType":89,"taxonomyTypeDesc":90},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":154,"categoryName":155,"taxonomyType":156,"taxonomyTypeDesc":157},"01JBSEEDCAT000000000000032","设计师","PROFESSION","适用职业",[146,159,160,25,23,22,161,19,162,163,164,165,166,167,168,169],"inclusion","yield","aggregation","proteins","solubility","proteinmpnn","ligandmpnn","ligand-aware","inverse-folding","expression","sequence-design",{"skillNo":171,"name":172,"nameZh":6,"description":173,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":174,"licenseFileUrl":6,"sourceHtmlUrl":175,"categories":176,"keywords":177,"publishedAt":31},"01M6DQ8W91A93MGGCEEW949HBC","scvelo","RNA velocity analysis with scVelo. Estimate cell state transitions from unspliced\u002Fspliced mRNA dynamics, infer trajectory directions, compute latent time, and identify driver genes in single-cell RNA-seq data. Complements Scanpy\u002FscVI-tools for trajectory inference.","BSD-3-Clause","https:\u002F\u002Fgithub.com\u002Fhuamu668\u002Fclawhub-openclaw-medical\u002Fblob\u002Fmain\u002Fscvelo\u002FSKILL.md",[],[172,178,179,180,181,182,183,184,185,186,187,22,137,188,189],"RNA","single-cell","infer","cell","transitions","time","compute","velocity","scanpy","analysis","state","inference"]