Loading BioAtlas...
Loading BioAtlas...
将生物大模型、对接计算引擎与组学管线并排横向对比。按获取方式、硬件开销及细分领域灵活筛选，在一分钟内得出客观可辩护的技术选型方案。
| 技术参数指标 | ＋ 添加工具 | |||
|---|---|---|---|---|
| Category | Biological Foundation Models | Biological AI Models | Biological AI Models | — |
| Subcategory | Protein Language Models | Protein Design | Protein Design | — |
| Company | Salesforce AI Research | Pacesa Lab | Boltz Team | — |
| Access mode | Open Source | Research code or publication | Research code or publication | — |
| Licence conditions | BSD-3-Clause | Check code and model weights separately | Check code and model weights separately | — |
| Compute requirements | 1x GPU (>= 16GB VRAM) for inference | Check selected checkpoint and runtime; no GPU catalog recommendation | Check selected checkpoint and runtime; no GPU catalog recommendation | — |
| Biological application | Enzyme engineering, antibody optimization, and zero-shot fitness prediction | 未明确记录 | 未明确记录 | — |
| Research workflow | Generate sequence variants with high predicted stability and activity | 未明确记录 | 未明确记录 | — |
| Evidence level | Peer-reviewed publication (NeurIPS 2023) | Peer-reviewed primary method | Primary research preprint | — |
| Commercial eligibility | Permissive open source release | Unknown, requires audit | Unknown, requires audit | — |
| Price | Free Open Source | Check selected model licence | Check selected model licence | — |
| Availability ⓘ | Available on GitHub and Hugging Face | Public research method/code | Public research method/code | — |
| Features | Trained on over 1 billion sequences from UniParc, BFD, and metagenomic sources, Controllable sequence generation with tag-based conditioning, High accuracy zero-shot variant effect prediction, Model sizes ranging from 150M to 6.4B parameters | AF2-guided optimization pipeline proposes protein binder designs for target structures., Candidate binder sequences, modeled protein complexes | Target-conditioned protein and peptide binder generation model as reported in its preprint., Binder sequence and structure candidates | — |
| Entry type | AI Model | Research AI model / method | Research AI model / method | — |
| AI role | Generative Protein Modeling | Published biomedical AI method | Published biomedical AI method | — |
| Input data | 未明确记录 | Target protein 3D coordinates; AF2/MPNN workflow and target interface constraints | Target context; design constraints | — |
| Output data | 未明确记录 | Candidate binder sequences; modeled protein complexes | Binder sequence and structure candidates | — |
| Validation | 未明确记录 | Requires independent dataset-specific evaluation | Requires independent dataset-specific evaluation | — |
| Type | Autoregressive protein language model | 未明确记录 | 未明确记录 | — |
| Intended use | 未明确记录 | 未明确记录 | 未明确记录 | — |
| Compatibility | 未明确记录 | 未明确记录 | 未明确记录 | — |
| Manufacturer | Salesforce AI Research | 未明确记录 | 未明确记录 | — |
| Integration evidence | https://github.com/salesforce/progen | 未明确记录 | 未明确记录 | — |
| Laboratory handoff | Sequences ready for oligonucleotide synthesis and screening | 未明确记录 | 未明确记录 | — |