Back to Details

esm

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

ESM: Evolutionary Scale Modeling

ESM：进化尺度建模

Overview

概述

ESM provides state-of-the-art protein language models for understanding, generating, and designing proteins. This skill enables working with two model families: ESM3 for generative protein design across sequence, structure, and function, and ESM C for efficient protein representation learning and embeddings.
ESM提供最先进的蛋白质语言模型，用于蛋白质的理解、生成与设计。该工具支持两大模型系列：用于跨序列、结构和功能生成式蛋白质设计的ESM3，以及用于高效蛋白质表示学习与嵌入的ESM C。

Core Capabilities

核心功能

1. Protein Sequence Generation with ESM3

1. 基于ESM3的蛋白质序列生成

Generate novel protein sequences with desired properties using multimodal generative modeling.
When to use:
  • Designing proteins with specific functional properties
  • Completing partial protein sequences
  • Generating variants of existing proteins
  • Creating proteins with desired structural characteristics
Basic usage:
python
from esm.models.esm3 import ESM3
from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
通过多模态生成建模，生成具有所需特性的新型蛋白质序列。
适用场景：
  • 设计具有特定功能特性的蛋白质
  • 补全部分蛋白质序列
  • 生成现有蛋白质的变体
  • 创建具有所需结构特征的蛋白质
基础用法：
python
from esm.models.esm3 import ESM3
from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig

Load model locally

本地加载模型

model: ESM3InferenceClient = ESM3.from_pretrained("esm3-sm-open-v1").to("cuda")
model: ESM3InferenceClient = ESM3.from_pretrained("esm3-sm-open-v1").to("cuda")

Create protein prompt

创建蛋白质提示

protein = ESMProtein(sequence="MPRT___KEND") # '_' represents masked positions
protein = ESMProtein(sequence="MPRT___KEND") # '_'代表掩码位置

Generate completion

生成补全序列

protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8)) print(protein.sequence)

**For remote/cloud usage via Forge API:**

```python
from esm.sdk.forge import ESM3ForgeInferenceClient
from esm.sdk.api import ESMProtein, GenerationConfig
protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8)) print(protein.sequence)

**通过Forge API实现远程/云端使用：**

```python
from esm.sdk.forge import ESM3ForgeInferenceClient
from esm.sdk.api import ESMProtein, GenerationConfig

Connect to Forge

连接到Forge

model = ESM3ForgeInferenceClient(model="esm3-medium-2024-08", url="https://forge.evolutionaryscale.ai", token="<token>")
model = ESM3ForgeInferenceClient(model="esm3-medium-2024-08", url="https://forge.evolutionaryscale.ai", token="<token>")

Generate

生成序列

protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))

See `references/esm3-api.md` for detailed ESM3 model specifications, advanced generation configurations, and multimodal prompting examples.
protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))

如需了解ESM3模型的详细规格、高级生成配置和多模态提示示例，请参阅`references/esm3-api.md`。

2. Structure Prediction and Inverse Folding

2. 结构预测与逆折叠

Use ESM3's structure track for structure prediction from sequence or inverse folding (sequence design from structure).
Structure prediction:
python
from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
使用ESM3的结构轨迹，从序列预测结构，或执行逆折叠（从结构设计序列）。
结构预测：
python
from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig

Predict structure from sequence

从序列预测结构

protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...") protein_with_structure = model.generate( protein, GenerationConfig(track="structure", num_steps=protein.sequence.count("_")) )
protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...") protein_with_structure = model.generate( protein, GenerationConfig(track="structure", num_steps=protein.sequence.count("_")) )

Access predicted structure

获取预测结构

coordinates = protein_with_structure.coordinates # 3D coordinates pdb_string = protein_with_structure.to_pdb()

**Inverse folding (sequence from structure):**

```python
coordinates = protein_with_structure.coordinates # 3D坐标 pdb_string = protein_with_structure.to_pdb()

**逆折叠（从结构生成序列）：**

```python

Design sequence for a target structure

为目标结构设计序列

protein_with_structure = ESMProtein.from_pdb("target_structure.pdb") protein_with_structure.sequence = None # Remove sequence
protein_with_structure = ESMProtein.from_pdb("target_structure.pdb") protein_with_structure.sequence = None # 移除现有序列

Generate sequence that folds to this structure

生成可折叠为该结构的序列

designed_protein = model.generate( protein_with_structure, GenerationConfig(track="sequence", num_steps=50, temperature=0.7) )
undefined
designed_protein = model.generate( protein_with_structure, GenerationConfig(track="sequence", num_steps=50, temperature=0.7) )
undefined

3. Protein Embeddings with ESM C

3. 基于ESM C的蛋白质嵌入

Generate high-quality embeddings for downstream tasks like function prediction, classification, or similarity analysis.
When to use:
  • Extracting protein representations for machine learning
  • Computing sequence similarities
  • Feature extraction for protein classification
  • Transfer learning for protein-related tasks
Basic usage:
python
from esm.models.esmc import ESMC
from esm.sdk.api import ESMProtein
生成高质量嵌入，用于下游任务如功能预测、分类或相似性分析。
适用场景：
  • 为机器学习提取蛋白质表示
  • 计算序列相似性
  • 为蛋白质分类提取特征
  • 为蛋白质相关任务进行迁移学习
基础用法：
python
from esm.models.esmc import ESMC
from esm.sdk.api import ESMProtein

Load ESM C model

加载ESM C模型

model = ESMC.from_pretrained("esmc-300m").to("cuda")
model = ESMC.from_pretrained("esmc-300m").to("cuda")

Get embeddings

获取嵌入

protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...") protein_tensor = model.encode(protein)
protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...") protein_tensor = model.encode(protein)

Generate embeddings

生成嵌入

embeddings = model.forward(protein_tensor)

**Batch processing:**

```python
embeddings = model.forward(protein_tensor)

**批量处理：**

```python

Encode multiple proteins

编码多个蛋白质

proteins = [ ESMProtein(sequence="MPRTKEIND..."), ESMProtein(sequence="AGLIVHSPQ..."), ESMProtein(sequence="KTEFLNDGR...") ]
embeddings_list = [model.logits(model.forward(model.encode(p))) for p in proteins]

See `references/esm-c-api.md` for ESM C model details, efficiency comparisons, and advanced embedding strategies.
proteins = [ ESMProtein(sequence="MPRTKEIND..."), ESMProtein(sequence="AGLIVHSPQ..."), ESMProtein(sequence="KTEFLNDGR...") ]
embeddings_list = [model.logits(model.forward(model.encode(p))) for p in proteins]

如需了解ESM C模型的详细信息、效率对比和高级嵌入策略，请参阅`references/esm-c-api.md`。

4. Function Conditioning and Annotation

4. 功能条件控制与注释

Use ESM3's function track to generate proteins with specific functional annotations or predict function from sequence.
Function-conditioned generation:
python
from esm.sdk.api import ESMProtein, FunctionAnnotation, GenerationConfig
使用ESM3的功能轨迹，生成具有特定功能注释的蛋白质，或从序列预测功能。
功能条件生成：
python
from esm.sdk.api import ESMProtein, FunctionAnnotation, GenerationConfig

Create protein with desired function

创建具有所需功能的蛋白质

protein = ESMProtein( sequence="_" * 200, # Generate 200 residue protein function_annotations=[ FunctionAnnotation(label="fluorescent_protein", start=50, end=150) ] )
protein = ESMProtein( sequence="_" * 200, # 生成200个残基的蛋白质 function_annotations=[ FunctionAnnotation(label="fluorescent_protein", start=50, end=150) ] )

Generate sequence with specified function

生成具有指定功能的序列

functional_protein = model.generate( protein, GenerationConfig(track="sequence", num_steps=200) )
undefined
functional_protein = model.generate( protein, GenerationConfig(track="sequence", num_steps=200) )
undefined

5. Chain-of-Thought Generation

5. 思维链生成

Iteratively refine protein designs using ESM3's chain-of-thought generation approach.
python
from esm.sdk.api import GenerationConfig
使用ESM3的思维链生成方法，迭代优化蛋白质设计。
python
from esm.sdk.api import GenerationConfig

Multi-step refinement

多步骤优化

protein = ESMProtein(sequence="MPRT" + "_" * 100 + "KEND")
protein = ESMProtein(sequence="MPRT" + "_" * 100 + "KEND")

Step 1: Generate initial structure

步骤1：生成初始结构

config = GenerationConfig(track="structure", num_steps=50) protein = model.generate(protein, config)
config = GenerationConfig(track="structure", num_steps=50) protein = model.generate(protein, config)

Step 2: Refine sequence based on structure

步骤2：基于结构优化序列

config = GenerationConfig(track="sequence", num_steps=50, temperature=0.5) protein = model.generate(protein, config)
config = GenerationConfig(track="sequence", num_steps=50, temperature=0.5) protein = model.generate(protein, config)

Step 3: Predict function

步骤3：预测功能

config = GenerationConfig(track="function", num_steps=20) protein = model.generate(protein, config)
undefined
config = GenerationConfig(track="function", num_steps=20) protein = model.generate(protein, config)
undefined

6. Batch Processing with Forge API

6. 基于Forge API的批量处理

Process multiple proteins efficiently using Forge's async executor.
python
from esm.sdk.forge import ESM3ForgeInferenceClient
import asyncio

client = ESM3ForgeInferenceClient(model="esm3-medium-2024-08", token="<token>")
使用Forge的异步执行器高效处理多个蛋白质。
python
from esm.sdk.forge import ESM3ForgeInferenceClient
import asyncio

client = ESM3ForgeInferenceClient(model="esm3-medium-2024-08", token="<token>")

Async batch processing

异步批量处理

async def batch_generate(proteins_list): tasks = [ client.async_generate(protein, GenerationConfig(track="sequence")) for protein in proteins_list ] return await asyncio.gather(*tasks)
async def batch_generate(proteins_list): tasks = [ client.async_generate(protein, GenerationConfig(track="sequence")) for protein in proteins_list ] return await asyncio.gather(*tasks)

Execute

执行

proteins = [ESMProtein(sequence=f"MPRT{'_' * 50}KEND") for _ in range(10)] results = asyncio.run(batch_generate(proteins))

See `references/forge-api.md` for detailed Forge API documentation, authentication, rate limits, and batch processing patterns.
proteins = [ESMProtein(sequence=f"MPRT{'_' * 50}KEND") for _ in range(10)] results = asyncio.run(batch_generate(proteins))

如需了解详细的Forge API文档、认证、速率限制和批量处理模式，请参阅`references/forge-api.md`。

Model Selection Guide

模型选择指南

ESM3 Models (Generative):
  • esm3-sm-open-v1
    (1.4B) - Open weights, local usage, good for experimentation
  • esm3-medium-2024-08
    (7B) - Best balance of quality and speed (Forge only)
  • esm3-large-2024-03
    (98B) - Highest quality, slower (Forge only)
ESM C Models (Embeddings):
  • esmc-300m
    (30 layers) - Lightweight, fast inference
  • esmc-600m
    (36 layers) - Balanced performance
  • esmc-6b
    (80 layers) - Maximum representation quality
Selection criteria:
  • Local development/testing: Use 
    esm3-sm-open-v1
    or 
    esmc-300m
  • Production quality: Use 
    esm3-medium-2024-08
    via Forge
  • Maximum accuracy: Use 
    esm3-large-2024-03
    or 
    esmc-6b
  • High throughput: Use Forge API with batch executor
  • Cost optimization: Use smaller models, implement caching strategies
ESM3生成式模型：
  • esm3-sm-open-v1
    （14亿参数） - 开源权重，本地使用，适合实验
  • esm3-medium-2024-08
    （70亿参数） - 质量与速度的最佳平衡（仅支持Forge）
  • esm3-large-2024-03
    （980亿参数） - 最高质量，速度较慢（仅支持Forge）
ESM C嵌入模型：
  • esmc-300m
    （30层） - 轻量级，推理速度快
  • esmc-600m
    （36层） - 性能均衡
  • esmc-6b
    （80层） - 表示质量最优
选择标准：
  • 本地开发/测试： 使用
    esm3-sm-open-v1
    或
    esmc-300m
  • 生产级质量： 通过Forge使用
    esm3-medium-2024-08
  • 最高精度： 使用
    esm3-large-2024-03
    或
    esmc-6b
  • 高吞吐量： 使用Forge API搭配批量执行器
  • 成本优化： 使用较小模型，实现缓存策略

Installation

安装

Basic installation:
bash
uv pip install esm
With Flash Attention (recommended for faster inference):
bash
uv pip install esm
uv pip install flash-attn --no-build-isolation
For Forge API access:
bash
uv pip install esm  # SDK includes Forge client
No additional dependencies needed. Obtain Forge API token at https://forge.evolutionaryscale.ai
基础安装：
bash
uv pip install esm
搭配Flash Attention（推荐，可提升推理速度）：
bash
uv pip install esm
uv pip install flash-attn --no-build-isolation
For API访问：
bash
uv pip install esm  # SDK包含Forge客户端
无需额外依赖。请访问https://forge.evolutionaryscale.ai获取Forge API令牌。

Common Workflows

常见工作流

For detailed examples and complete workflows, see 
references/workflows.md
which includes:
  • Novel GFP design with chain-of-thought
  • Protein variant generation and screening
  • Structure-based sequence optimization
  • Function prediction pipelines
  • Embedding-based clustering and analysis
如需详细示例和完整工作流，请参阅
references/workflows.md
，其中包含：
  • 基于思维链的新型GFP设计
  • 蛋白质变体生成与筛选
  • 基于结构的序列优化
  • 功能预测流水线
  • 基于嵌入的聚类与分析

References

参考资料

This skill includes comprehensive reference documentation:
  • references/esm3-api.md
    - ESM3 model architecture, API reference, generation parameters, and multimodal prompting
  • references/esm-c-api.md
    - ESM C model details, embedding strategies, and performance optimization
  • references/forge-api.md
    - Forge platform documentation, authentication, batch processing, and deployment
  • references/workflows.md
    - Complete examples and common workflow patterns
These references contain detailed API specifications, parameter descriptions, and advanced usage patterns. Load them as needed for specific tasks.
本工具包含全面的参考文档：
  • references/esm3-api.md
    - ESM3模型架构、API参考、生成参数和多模态提示
  • references/esm-c-api.md
    - ESM C模型细节、嵌入策略和性能优化
  • references/forge-api.md
    - Forge平台文档、认证、批量处理和部署
  • references/workflows.md
    - 完整示例和常见工作流模式
这些参考资料包含详细的API规范、参数说明和高级使用模式。可根据具体任务按需加载。

Best Practices

最佳实践

For generation tasks:
  • Start with smaller models for prototyping (
    esm3-sm-open-v1
    )
  • Use temperature parameter to control diversity (0.0 = deterministic, 1.0 = diverse)
  • Implement iterative refinement with chain-of-thought for complex designs
  • Validate generated sequences with structure prediction or wet-lab experiments
For embedding tasks:
  • Batch process sequences when possible for efficiency
  • Cache embeddings for repeated analyses
  • Normalize embeddings when computing similarities
  • Use appropriate model size based on downstream task requirements
For production deployment:
  • Use Forge API for scalability and latest models
  • Implement error handling and retry logic for API calls
  • Monitor token usage and implement rate limiting
  • Consider AWS SageMaker deployment for dedicated infrastructure
生成任务：
  • 原型开发从较小模型开始（
    esm3-sm-open-v1
    ）
  • 使用temperature参数控制多样性（0.0=确定性，1.0=高多样性）
  • 对复杂设计使用思维链进行迭代优化
  • 通过结构预测或湿实验验证生成的序列
嵌入任务：
  • 尽可能批量处理序列以提升效率
  • 为重复分析缓存嵌入结果
  • 计算相似性时对嵌入进行归一化
  • 根据下游任务需求选择合适的模型大小
生产部署：
  • 使用Forge API实现可扩展性和获取最新模型
  • 为API调用实现错误处理和重试逻辑
  • 监控令牌使用情况并实现速率限制
  • 考虑使用AWS SageMaker部署专用基础设施

Resources and Documentation

资源与文档

Responsible Use

负责任使用

ESM is designed for beneficial applications in protein engineering, drug discovery, and scientific research. Follow the Responsible Biodesign Framework (https://responsiblebiodesign.ai/) when designing novel proteins. Consider biosafety and ethical implications of protein designs before experimental validation.
ESM专为蛋白质工程、药物发现和科学研究中的有益应用而设计。设计新型蛋白质时，请遵循《负责任生物设计框架》(https://responsiblebiodesign.ai/)。在实验验证前，请考虑蛋白质设计的生物安全和伦理影响。