# AI use cases for Colab Fold (sokrypton/ColabFold)

Repo: [sokrypton/ColabFold](https://github.com/sokrypton/ColabFold) | Use cases: 3 | Page: https://dedupedai.com/use-cases/sokrypton-colabfold/

ColabFold makes protein structure prediction accessible through Google Colab notebooks for AlphaFold2 with MMseqs2, AlphaFold3 (OpenFold3), ESMFold, batch runs and beta models such as RoseTTAFold2, Boltz and BioEmu. It also runs locally: colabfold_search builds multiple sequence alignments against the ColabFold databases (with optional GPU-accelerated MMseqs2-GPU search) and colabfold_batch predicts structures, and both can export MSAs as AlphaFold3-compatible JSON.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Fold a protein in a free Colab notebook: https://dedupedai.com/use-cases/sokrypton-colabfold/fold-a-protein-in-a-colab-notebook/
- Batch-fold many proteins on your own machines: https://dedupedai.com/use-cases/sokrypton-colabfold/batch-fold-many-proteins-locally-with-colabfold/
- Prepare AlphaFold3 inputs with Colab Fold alignments: https://dedupedai.com/use-cases/sokrypton-colabfold/prepare-alphafold3-inputs-with-colabfold-msas/

## How to fold a protein in a free Colab notebook with Colab Fold

Repo: [sokrypton/ColabFold](https://github.com/sokrypton/ColabFold) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README

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

**Who it is for:** A biology student or wet-lab researcher who needs a structure model for one protein or complex today.

**The problem:** Running AlphaFold locally needs big databases, a GPU and setup time most people do not have.

**The outcome:** A predicted structure with pLDDT confidence values, from a notebook in your browser.

### Steps

1. Open the AlphaFold2_mmseqs2 notebook from the table in the README (monomers and complexes, MMseqs2 MSAs, templates)
2. Paste your sequence and run the cells; use ESMFold or the AlphaFold3 (OpenFold3) notebook if you want to compare
3. Download the predicted structure files when the run finishes
4. Color by pLDDT in PyMOL with `spectrum b, red_yellow_green_cyan_blue, minimum=50, maximum=90`
5. For molecular replacement, remember the bfactor column holds pLDDT (higher is better), not a real B-factor

### Prompt to paste into your agent

```text
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.
```

### Caveat

Maximum length depends on the free GPU Google Colab gives you; the README suggests about 2000 residues on 16 GB GPUs.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/sokrypton-colabfold/fold-a-protein-in-a-colab-notebook/

## How to batch-fold many proteins on your own machines with Colab Fold

Repo: [sokrypton/ColabFold](https://github.com/sokrypton/ColabFold) | Difficulty: Advanced | Time: half a day | Basis: From the README

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

**Who it is for:** A lab with a list of hundreds of sequences that would take days to paste into notebooks.

**The problem:** Notebooks are one sequence at a time and the public MSA server asks you not to hammer it from many machines.

**The outcome:** An msas folder of a3m alignments and a predictions folder with pdb, json and png files for every sequence.

### Steps

1. Set up a local Colab Fold install, MMseqs2 and the Colab Fold databases with setup_databases.sh
2. Build alignments (CPU heavy) with `colabfold_search --mmseqs /path/to/bin/mmseqs input_sequences.fasta /path/to/db_folder msas`
3. Predict structures (needs a GPU) with `colabfold_batch msas predictions`
4. Review the predicted pdb, json and png files in the predictions folder
5. Keep MMSEQS_NO_INDEX=1 for most uses; build the precomputed index only for a server or very fast storage

### Prompt to paste into your agent

```text
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.
```

### Caveat

Local installation steps are not in the README text itself; use the project's linked install instructions.

Risk, platform: The databases are very large and the search needs a lot of CPU; fast single-query searches with a full in-memory index need 768GB to 1TB of RAM.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/sokrypton-colabfold/batch-fold-many-proteins-locally-with-colabfold/

## How to prepare AlphaFold3 inputs with Colab Fold alignments with Colab Fold

Repo: [sokrypton/ColabFold](https://github.com/sokrypton/ColabFold) | Difficulty: Advanced | Time: 1 hour | Basis: From the README

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

**Who it is for:** A structural biologist running AlphaFold3 who wants Colab Fold's fast MMseqs2 alignments instead of AF3's own search.

**The problem:** Generating MSAs inside AlphaFold3 is slow, and hand-writing its JSON for complexes with ligands is error prone.

**The outcome:** AlphaFold3 input JSON files, one per entry, ready to feed to AlphaFold3.

### Steps

1. Write complexes in FASTA, adding non-protein parts as molecule type|sequence|(copies), for example ccd|ATP|2 or dna|ATCG
2. With local databases run `colabfold_search --mmseqs /path/to/bin/mmseqs input_sequences.fasta /path/to/db_folder msas --af3-json`
3. Or with the MSA server run `colabfold_batch input_sequences.fasta out_dir --af3-json` (prediction is skipped, only JSON is written)
4. Speed up searches on GPUs: set up GPU databases with `GPU=1 ./setup_databases.sh /path/to/db_folder` and add --gpu 1
5. Pass the generated JSON files to AlphaFold3

### Prompt to paste into your agent

```text
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 ';'.
```

### Caveat

MMseqs2 MSAs are only generated for protein sequences; RNA entries get a null unpaired MSA so AlphaFold3 builds its own.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/sokrypton-colabfold/prepare-alphafold3-inputs-with-colabfold-msas/