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 AI Models | Biological Foundation Models | Biological AI Models | — |
| Subcategory | Protein Structure | Multimodal Generative Models | Protein Structure | — |
| Company | Google DeepMind | EvolutionaryScale | Baker laboratory | — |
| Access mode | Provider-specific terms | Open weights & Cloud API | Public code / project terms | — |
| Licence conditions | Code and model parameters have separate terms. Parameter access must be obtained from the provider; commercial eligibility is not established here. | EvolutionaryScale Open Model License (1.4B) / Commercial Cloud API | Review the linked release, model weights, datasets and service terms separately | — |
| Compute requirements | H100 80 GB; CPU/RAM and sequence databases are also required. Larger inputs can require configuration changes. | 1x A100 GPU (1.4B open model) to multi-node H100 clusters (98B model) | Consult linked installation guidance; workload dependent | — |
| Biological application | 未明确记录 | De novo protein generation, motif scaffolding, and functional sequence design | 未明确记录 | — |
| Research workflow | Predict biomolecular structures | Structure & function generation from biological prompts | 未明确记录 | — |
| Evidence level | Provider-tested configuration | Peer-reviewed research and experimental wet-lab validation (Science 2024) | Project or provider documentation; no local performance benchmark | — |
| Commercial eligibility | Not independently confirmed | Commercial API & enterprise licenses available; non-commercial for open weights | Not independently confirmed | — |
| Price | See provider terms | Free open weights (1.4B) / Paid enterprise API | See project terms; compute costs may apply | — |
| Availability ⓘ | Check provider access requirements | Globally available via AWS Bedrock and EvolutionaryScale API | See project access and maintenance status | — |
| Features | Biomolecular structure prediction, Local inference pipeline | Simultaneous multimodal generation across sequence, coordinates, and secondary structure, Promptable with structural motifs, active site geometries, or sequence fragments, Demonstrated de novo generation of fluorescent proteins with 500M years of divergent evolution, Available in 1.4B open weights, 7B, and 98B parameter frontier configurations | Structure modeling of proteins with molecular partners, Input: Sequences and configured nucleic-acid or small-molecule inputs, Output: Predicted biomolecular complex coordinates | — |
| Entry type | Software / platform | AI Model | Research software / model | — |
| AI role | AI / computational workflow | Multimodal Generative AI | Scientific machine learning | — |
| Input data | 未明确记录 | 未明确记录 | Sequences and configured nucleic-acid or small-molecule inputs | — |
| Output data | 未明确记录 | 未明确记录 | Predicted biomolecular complex coordinates | — |
| Validation | 未明确记录 | 未明确记录 | Evaluate on representative data; no clinical suitability established | — |
| Type | Structure prediction model | Generative protein language & structure model | 未明确记录 | — |
| Intended use | 未明确记录 | 未明确记录 | Biological and pharmaceutical research | — |
| Compatibility | 未明确记录 | 未明确记录 | 未明确记录 | — |
| Access ⓘ | See code licence and model parameter terms | 未明确记录 | 未明确记录 | — |
| Integration evidence | https://github.com/google-deepmind/alphafold3/blob/main/docs/performance.md | https://www.evolutionaryscale.ai/blog/esm-3-release | 未明确记录 | — |
| Laboratory handoff | Editorial next step: select an appropriate structural or functional validation assay with the research team. No direct instrument integration has been verified. | Output FASTA sequences can be sent directly to gene synthesis vendors (Twist, IDT) | 未明确记录 | — |
| Manufacturer | 未明确记录 | EvolutionaryScale | 未明确记录 | — |
| Limitations | 未明确记录 | 未明确记录 | Correct chemical inputs and model-specific preprocessing are essential; binding is not experimentally established. | — |