sokrypton/ColabDesign

What it solves

It makes protein design accessible to a wider audience by providing a suite of tools integrated into Google Colab, removing the need for complex local installations of specialized software.

How it works

The project provides a set of notebooks that implement different protein design paradigms based on probability distributions:

• Structure from Sequence: Uses TrRosetta (TrDesign) and AlphaFold (AfDesign) to predict structure from sequence.
• Sequence from Structure: Uses ProteinMPNN to determine the sequence that fits a specific structure.
• Sequence Generation: Implements tools like MSA_transformer and various DCA methods (GREMLIN, mfDCA, etc.) to generate sequences.
• Structure Generation: Uses Rfdiffusion to generate new protein structures.

Who it’s for

Researchers and scientists who need to design proteins but may not have the high-end computing resources or the expertise to set up complex bioinformatics software locally.

Highlights

• Integration with Google Colab for easy access.
• Support for multiple design methods including AlphaFold and RoseTTAFold.
• Support for sequence-to-structure and structure-to-sequence design.
• Includes Rfdiffusion for structure generation.