Can I use ESM models locally or do I need an API key?

You can use the skill locally with open-weight models like esm3-sm-open-v1. For larger production-grade models, you can connect to the Forge API using an API token from EvolutionaryScale.

Does this skill support structure prediction from sequence?

Yes, the skill utilizes ESM3's structure track to predict 3D coordinates from a protein sequence, and it can also perform inverse folding to design sequences for a target structure.

What is the difference between ESM3 and ESM C in this skill?

ESM3 is a multimodal generative model used for creating new protein sequences, structures, and functions, whereas ESM C is optimized for creating efficient, high-quality embeddings and representations for downstream analysis.

How do I install the necessary dependencies for this skill?

The core library can be installed via 'uv pip install esm'. For improved performance, installing flash-attention is recommended for local GPU-based inference.

Protein Modeling with ESM

Name: Protein Modeling with ESM
Author: ovachiever

byovachiever

•

Data Science & ML

Provides advanced protein language model capabilities for sequence generation, structure prediction, and functional design using ESM3 and ESM C.

About

The ESM skill integrates state-of-the-art protein language models into your workflow, enabling researchers and developers to perform complex computational biology tasks directly within Claude Code. It features ESM3 for generative protein design across sequence, structure, and function, alongside ESM C for efficient representation learning. Whether you are designing novel proteins with specific functional properties, predicting 3D structures from sequences, or generating high-quality embeddings for machine learning classification, this skill provides a standardized interface for both local execution and scalable cloud-based inference via the Forge API.

Key Features

Multimodal generative protein design (sequence, structure, and function)
3D structure prediction and inverse folding capabilities
High-quality protein embeddings using ESM C models up to 6B parameters
15 GitHub stars
Function-conditioned generation and automated annotation
Support for local inference and cloud-based Forge API scaling

Use Cases

Designing novel therapeutic proteins with specific structural or functional motifs
Refining protein variants through iterative chain-of-thought design cycles
Generating high-dimensional embeddings for protein similarity and function prediction

About

Key Features

Multimodal generative protein design (sequence, structure, and function)
3D structure prediction and inverse folding capabilities
High-quality protein embeddings using ESM C models up to 6B parameters
15 GitHub stars
Function-conditioned generation and automated annotation
Support for local inference and cloud-based Forge API scaling

Use Cases

Designing novel therapeutic proteins with specific structural or functional motifs
Refining protein variants through iterative chain-of-thought design cycles
Generating high-dimensional embeddings for protein similarity and function prediction