How to use Colab Fold: what you can do with it

3 use cases

• Fold a protein in a free Colab notebook

  Open the AlphaFold2_mmseqs2 notebook, paste a sequence and get a predicted structure without installing anything.

  • REPO sokrypton/ColabFold
  • Security and science
  • Beginner
  View prompt

  I want to use sokrypton/ColabFold's AlphaFold2_mmseqs2 notebook for this sequence: <sequence>. Walk me through the settings to choose, then explain how to read the pLDDT values in the result.

  Share Email me this

• Batch-fold many proteins on your own machines

  Run colabfold_search for alignments and colabfold_batch for predictions over a whole FASTA file.

  • REPO sokrypton/ColabFold
  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • Security and science
  • Advanced
  View prompt

  Using sokrypton/ColabFold locally, run colabfold_search on <fasta file> against my databases at <db folder>, then colabfold_batch on the msas folder. Summarise the mean pLDDT per sequence from the outputs.

  Share Email me this

• Prepare AlphaFold3 inputs with Colab Fold alignments

  Export Colab Fold MSAs as AlphaFold3-compatible JSON, including ligands and nucleic acids written in the FASTA.

  • REPO sokrypton/ColabFold
  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • Security and science
  • Advanced
  View prompt

  With sokrypton/ColabFold, turn <my FASTA with proteins and ligands> into AlphaFold3 input JSON using --af3-json. Check the FASTA uses the molecule type|sequence|(copies) format and replace aromatic ':' bonds in SMILES with ';'.

  Share Email me this

Quick start with Colab Fold

1. Check you have what it needs: A Google account for the Colab notebooks, or a local ColabFold install for colabfold_search and colabfold_batch; For local searches: MMseqs2 and the ColabFold databases (large downloads, a lot of CPU); A GPU for structure prediction; about 16 GB of GPU memory handles sequences up to roughly 2000 residues.
2. Open sokrypton/ColabFold on GitHub and read the README and license.
3. Open the AlphaFold2_mmseqs2 notebook from the table in the README (monomers and complexes, MMseqs2 MSAs, templates)
4. Paste your sequence and run the cells; use ESMFold or the AlphaFold3 (OpenFold3) notebook if you want to compare
5. Download the predicted structure files when the run finishes
6. Color by pLDDT in PyMOL with spectrum b, red_yellow_green_cyan_blue, minimum=50, maximum=90

These steps come from our top use case, Fold a protein in a free Colab notebook, which also has a copy-paste agent prompt.

Using the Deduped MCP, get the setup steps for sokrypton/ColabFold

claude mcp add --transport http deduped https://dedupedai.com/mcp

Best AI repos for Security and science

About this repo

• Who it is forResearchers and students who want protein folding without their own cluster, and labs that run many predictions locally.
• Why it is hot3k stars, talked about in 2 AI sources we track
• LanguageJupyter Notebook
• Sourcegithub.com/sokrypton/ColabFold · Back to Hot repos

Requirements

• A Google account for the Colab notebooks, or a local ColabFold install for colabfold_search and colabfold_batch
• For local searches: MMseqs2 and the ColabFold databases (large downloads, a lot of CPU)
• A GPU for structure prediction; about 16 GB of GPU memory handles sequences up to roughly 2000 residues
• For GPU search: GPU databases set up with GPU=1

Related repos

• Trended earlier REPO google-deepmind/alphafold3 Alphafold3 AlphaFold 3 is Google DeepMind's inference pipeline for predicting the structure of biomolecular complexes. You describe the complex in a JSON input file and run run_alphafold.py, usually inside the provided container, with separate flags for the CPU heavy data pipeline and the GPU inference step. Model parameters are not in the repo: you must request them from Google and follow their terms of use.
• Trended earlier REPO jwohlwend/boltz Boltz Boltz is a family of open models for biomolecular interaction prediction. Boltz-1 was the first fully open source model to approach AlphaFold3 accuracy, and Boltz-2 jointly models complex structures and binding affinities, approaching the accuracy of physics-based free-energy perturbation (FEP) methods while running about 1000x faster. Code and weights are MIT licensed for academic and commercial use, and the boltz command runs predictions from YAML input files.
• Trended earlier REPO Biohub/esm 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.