New Bindcraft tutorial now available

BindCraft is a new open-source pipeline for de novo protein binder design developed by Martin Pacesa in collaboration with the Correia Lab (EPFL) and the Ovchinnikov Lab (MIT). Designed to be accessible and efficient, BindCraft was recently released with a detailed tutorial and can be run locally or via Google Colab. The tool enables researchers to generate protein binders entirely from scratch using only the structure of a target protein.

“We’ve written up a tutorial on how to run BindCraft, how to prepare your input PDB, how to select hotspots, and various other tips and tricks to get the most out of binder design!” – Martin Pacesa.

How This Document Helps

This new write-up provides important context that enhances usability of the tool when combined with the information provided in the README. It provides more detailed explanations for the use cases of BindCraft, how to analyze output structures, and what is happening ‘under the hood.’ Users running BindCraft for the first time will find it a helpful companion to the technical documentation.

What the Guide Covers

The write-up outlines key components and design choices within BindCraft, including:

• Installation: BindCraft only runs on Linux-based systems and can be installed with the Conda or Mamba package managers. To use BindCraft, a license for PyRosetta is necessary. 
• Hardware considerations: BindCraft requires a CUDA-compatible NVIDIA GPU or can be run via a CoLab notebook. The tool is memory-intensive; having access to a few terabytes of storage is recommended.
• Getting started: Gives details about the information needed by BindCraft to generate designs, the types of biological situations the tool works best for, and recommendations for how to create the most successful binder designs. 
• Under the hood: A brief description of the workflow includes backpropagation via AlphaFold2-Multimer, redesign with ProteinMPNN, and final filtering using scoring metrics.
• Target preparation & Hotspot selection: Explains how to select a suitable target chain and how to define hotspot residues that guide binder interaction.
• Advanced settings: Covers optional parameters, such as controlling binder length and secondary structure bias, which can influence the success rate and design diversity.
• Filtering: Details post-run filters, such as AF2 score thresholds and Rosetta ddG scoring, to identify promising binder candidates.
• Design selection: Discusses strategies for choosing final designs, such as ranking by score or selecting diverse top performers.
• Troubleshooting & Common Questions: Addresses common issues, such as poor success rates, and guides interpreting and adjusting inputs to improve design outcomes.

Access the Guide

The full guide is available on the BindCraft GitHub Wiki. For users interested in customizing their binder design runs, particularly in terms of input setup and result interpretation, this resource offers detailed, experience-informed guidance.