DEDUPEDWhat do you want to do?
3 use cases
-
Turn protein sequences into embeddings with ESMC
Load ESMC 6B from Hugging Face with the Esm package and get per layer representations for your sequences.
View prompt
Using Biohub/esm, write a script that installs esm with `pip install esm`, loads ESMC-6B with EsmcForMaskedLM.from_pretrained, and saves all layer hidden states for every sequence in <my FASTA file> as a NumPy array.
-
Predict an all atom structure from a sequence
Run ESMFold2 locally to fold a protein, including DNA chains and ligands, and write the result as mmCIF.
View prompt
With Biohub/esm, fold <my protein sequence> with ESMFold2 using ESMFold2InputBuilder().fold, print pLDDT, pTM and ipTM, and save the structure as a mmCIF file I can open in a viewer.
-
Read which learned features a protein activates
Attach an ESMC sparse autoencoder and inspect the sparse feature activations for a sequence.
View prompt
Using Biohub/esm, attach the ESMC-6B SAE to layers 30 and 60, run <my sequence> through it, and list the strongest sparse features with their shapes.
What people are building with Esm
-
Benchmarkthe web
Steered ESMFold2 binder designs toward beta-sheets with a helix penalty
Arcadia Science noticed ESMFold2 produced designs with more alpha-helical content than natural proteins, so they added a penalty for helix-like geometry during binder design without retraining the model. Across five targets the designs shifted toward beta-sheet and loop content, with lower average confidence, and they published the code, predicted structures and optimization traces.
By Arcadia Science (thestacks.org) on the web
Real builds shared on X, Hacker News, Reddit, DEV and GitHub, summed up by Deduped from each post and its links; each card links to the original post. See what people are building with other repos
Quick start with Esm
- Check you have what it needs: Python with pip; `pip install esm` is about 5.6 GB, of which about 4.6 GB is the CUDA build of torch; A CUDA GPU for the local examples, which load models with device set to cuda; A Biohub Platform API token if you use the hosted client instead of local weights.
- Open Biohub/esm on GitHub and read the README and license.
- Install the package with
pip install esm - Load the model and tokenizer with EsmcForMaskedLM.from_pretrained and EsmcTokenizer as in the README example
- Tokenize your sequences with return_tensors set to pt and padding set to True, then move the inputs to model.device
- Run the model inside torch.inference_mode, and pass output_hidden_states=True to get hidden states from all transformer layers
These steps come from our top use case, Turn protein sequences into embeddings with ESMC, which also has a copy-paste agent prompt.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for Biohub/esm
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeAbout this repo
- Who it is forComputational biologists and ML engineers who want protein embeddings, structure predictions or interpretable features without building the stack themselves.
- Why it is hot3k stars, picked up by one AI source we track
- Sourcegithub.com/Biohub/esm ยท Back to Hot repos
Requirements
- Python with pip; `pip install esm` is about 5.6 GB, of which about 4.6 GB is the CUDA build of torch
- A CUDA GPU for the local examples, which load models with device set to cuda
- A Biohub Platform API token if you use the hosted client instead of local weights
Add this badge to your README
Maintain Esm? Show readers where to find these use cases. 
Markdown
[](https://dedupedai.com/use-cases/biohub-esm/?utm_source=github&utm_medium=badge&utm_campaign=Biohub-esm)
HTML
<a href="https://dedupedai.com/use-cases/biohub-esm/?utm_source=github&utm_medium=badge&utm_campaign=Biohub-esm"><img src="https://dedupedai.com/badge/Biohub/esm.svg" alt="Use cases on Deduped" height="20"></a>
A static image, no tracking script. About the badge
Questions about Esm
What is Esm?
The esm package from Biohub (EvolutionaryScale) ships ESMC, a protein language model trained on billions of protein sequences, ESMFold2, a structure prediction model built on ESMC 6B that predicts all atom 3D structures from sequence, and sparse autoencoders that decompose ESMC representations into interpretable features. Models run locally from Hugging Face weights or through the Biohub Platform with an API token. It is an open-source project on GitHub: https://github.com/Biohub/esm.
How do I use Esm?
Install the package with pip install esm; then load the model and tokenizer with EsmcForMaskedLM.from_pretrained and EsmcTokenizer as in the README example; Tokenize your sequences with return_tensors set to pt and padding set to True, then move the inputs to model.device. The use cases on this page give the full steps and a copy-paste agent prompt for each job.
Who is Esm for?
Computational biologists and ML engineers who want protein embeddings, structure predictions or interpretable features without building the stack themselves.
What can I do with Esm?
3 use cases so far, for example: Turn protein sequences into embeddings with ESMC; Predict an all atom structure from a sequence; Read which learned features a protein activates. Each has steps and a copy-paste agent prompt.
What do I need to run Esm?
Python with pip; `pip install esm` is about 5.6 GB, of which about 4.6 GB is the CUDA build of torch; A CUDA GPU for the local examples, which load models with device set to cuda; A Biohub Platform API token if you use the hosted client instead of local weights.
What are alternatives to Esm?
Other hot AI repos in Security and science: Security Audit Skill, GPU Unlock, Syft, Alphafold3, Osv Scanner.
Is Esm safe to run?
Deduped does not own, audit or vouch for it. Read the README and license, run it on your own systems first and review every command before you run it.