DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

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

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

Short answer: Write complexes in FASTA, adding non-protein parts as molecule type|sequence|(copies), for example ccd|ATP|2 or dna|ATCG; then with local databases run colabfold_search --mmseqs /path/to/bin/mmseqs input_sequences.fasta /path/to/db_folder msas --af3-json; Or with the MSA server run colabfold_batch input_sequences.fasta out_dir --af3-json (prediction is skipped, only JSON is written). All steps.

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

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

The outcomeAlphaFold3 input JSON files, one per entry, ready to feed to AlphaFold3.

What you need first

  • 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

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 for your agent

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.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for sokrypton/ColabFold and "Prepare AlphaFold3 inputs with Colab Fold alignments"

Add Deduped to Claude Code once

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

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer