sokrypton/ColabFold

Making Protein folding accessible to all!

What it solves

ColabFold is a tool designed to make high-accuracy protein structure prediction accessible to all users by providing easy-to-use Google Colab notebooks and a local installation option. It removes the need for complex local setup of massive databases and high-end hardware by offering a public MSA (Multiple Sequence Alignment) server, a batch processing tool, and support for multiple state-of-the-art protein folding models.

How it works

ColabFold integrates several protein folding models (such as AlphaFold2, AlphaFold3/OpenFold3, ESMFold, and RoseTTAFold2) into a unified workflow. It uses an external MSA server to generate the sequence alignments required for folding predictions, which significantly reduces the local storage and compute requirements. For local users, it provides colabfold_batch for structure prediction and colabfold_search for local MSA generation using MMseqs2, including support for GPU-accelerated searches.

Who it’s for

Biologists, biochemists, and structural biologists who need to predict the 3D structure of proteins, monomers, and complexes without managing massive genomic databases or complex infrastructure.

Highlights

  • Multi-model support: Provides notebooks and tools for AlphaFold2, AlphaFold3, ESMFold, and RoseTTAFold2.
  • MSA Server: A public server that handles the computationally expensive part of sequence alignment, making folding accessible via Colab.
  • Local Installation: Supports Linux, macOS, and Windows (via WSL2) for high-throughput batch predictions.
  • GPU Acceleration: Supports GPU-accelerated MSA searches via MMseqs2-GPU and fused kernels for faster folding speed.
  • AF3 Compatibility: Ability to export MSAs into JSON format compatible with AlphaFold3, including support for non-protein molecules like DNA, RNA, and ligands.

Related

  • Project
    sokrypton/ColabDesign

    A collection of Google Colab notebooks that make protein design accessible by integrating tools like AlphaFold, ProteinMPNN, and Rfdiffusion.

  • Project
    aqlaboratory/openfold-3

    An open-source reproduction of AlphaFold3 for predicting the 3D structures of proteins, RNA, DNA, and small molecules.

  • Dispatch
    AlphaFold: Five Years of Scientific Impact

    Google DeepMind's AlphaFold has transformed biological research by solving the 50-year-old protein structure prediction problem, enabling over 3 million researchers to accelerate drug discovery and disease understanding.

  • Project
    google-deepmind/alphafold3

    AlphaFold 3 is an AI system for predicting the 3D structures of biomolecular interactions, helping researchers model how proteins and other molecules interact.

  • Project
    bytedance/Protenix

    Protenix is an open‑source deep‑learning system for high‑accuracy protein (and RNA‑protein) structure prediction. It builds on AlphaFold 3, offers several model sizes (including a fast Mini version), supports MSA/template/RNA inputs, and adds constraint‑aware folding. The package is installable via pip, includes a CLI, and a public web server. Benchmarks claim it outperforms AlphaFold 3 on antibody‑antigen complexes. All code is Apache 2.0; only the newest v2 weights are proprietary.