Introducing Chaperone: A Unified Framework for Protein Design Workflows

Protein design workflows bring together multiple tools for generating protein structures, designing sequences, predicting structures, and evaluating candidate designs. Connecting these tools into a coherent pipeline can require substantial work: managing inputs and outputs, configuring individual tools, coordinating computational jobs, and tracking results across multiple stages.

Chaperone, a new tool in the Rosetta Commons GitHub organization, aims to simplify this process by providing a unified framework for building, executing, and evaluating structural biology workflows.

Developed by members of the Institute for Protein Design at the University of Washington, Chaperone makes it easier to combine existing tools into customizable, reproducible pipelines, helping researchers move from individual modeling tools to integrated protein design workflows.

It provides integrations for a range of structural biology and protein design tools, with documented blocks for components including RFdiffusion3, AlphaFold3, ProteinMPNN, and Boltz-2. The framework also supports additional tools and custom blocks, allowing researchers to extend workflows to accommodate their own methods and computational requirements.

Building workflows from composable blocks

Rather than treating a protein design workflow as a single, monolithic script, Chaperone organizes it into blocks: individual components that perform specific tasks and can be combined into larger pipelines.

These blocks can represent different stages of a design workflow, including:

• Backbone generation: Generate candidate protein structures using tools such as RFdiffusion3.
• Sequence design: Design amino acid sequences compatible with generated structures using tools such as ProteinMPNN.
• Structure prediction: Predict structures using tools such as AlphaFold3 or Boltz-2.
• Evaluation and filtering: Assess candidate designs using structural metrics and filter out designs that do not meet specified criteria.

By connecting these components, researchers can construct workflows tailored to their particular design goals. Chaperone also supports more advanced patterns, including branching workflows, parallel job submission, and iterative cycles that feed results from one stage back into another.

This modular approach makes it easier to adapt workflows as research needs change, without having to rebuild the entire pipeline each time a tool or design strategy changes.

Automating execution and improving reproducibility

Building a workflow is only part of the challenge. Running it reliably across multiple tools and computational jobs introduces additional complexity.

Chaperone provides automatic job orchestration to coordinate workflow execution, helping manage the dependencies between stages. Its framework also supports data lineage tracking and reproducible execution, making it easier to understand how outputs were generated and how different stages of a pipeline relate to one another.

These capabilities are particularly useful for workflows involving multiple computational tools, where keeping track of intermediate results, configurations, and job execution can otherwise become cumbersome.

By bringing these responsibilities into a shared framework, Chaperone helps researchers focus more on designing and evaluating proteins and less on the infrastructure required to connect their tools.

Documentation and agent-assisted workflows

Getting started with a new framework can be challenging, particularly when it needs to coordinate multiple tools, containers, and computational environments.

Chaperone’s documentation provides several resources to help users get started and explore its capabilities:

• Simple Example: Walk through a minimal pipeline connecting RFdiffusion3, ProteinMPNN, and AlphaFold3.
• All Blocks: Explore available blocks and their configuration parameters.
• Syntax Reference: Learn about the core APIs used to define and execute workflows.
• Validation: Explore functionality for evaluating structure prediction tools.

Chaperone also includes agent skills intended to help guide users through common tasks. These provide another way to interact with the framework and its setup process.

Together, these resources offer a starting point for researchers who want to build their first pipeline as well as those looking to develop more sophisticated, customized workflows.

Explore Chaperone

Chaperone offers a framework for bringing structural biology tools together into flexible, coordinated workflows. By combining modular components with job orchestration, data lineage tracking, and extensive documentation, it aims to make complex protein design pipelines easier to build, execute, and reproduce.

Whether you are developing a new design workflow or looking to connect tools you already use, Chaperone is now available to add to your toolkit..

Get started:

• GitHub repository: RosettaCommons/chaperone
• Documentation: Chaperone Documentation

Featured image inspired by PDB 2EZB.