PacesaLab/BindCraft2

What it solves

BindCraft2 (BC2) enables the computational design of protein binders—such as de novo miniproteins, scaffolded binders, and cyclic peptides—tailored to specific biological targets. It streamlines the process of creating proteins that can bind to a target molecule with desired properties, reducing the need for trial-and-error in the lab by providing ranked candidates for experimental validation.

How it works

BC2 uses a multi-stage workflow that combines several AI models and structural filters:

1. Sequence Optimisation: It utilizes AlphaFold 2 to optimize the protein sequence.
2. Redesign: It employs ProteinMPNN for sequence redesign.
3. Evaluation: Candidates are evaluated using separate AlphaFold models and structural filters to measure interface confidence, fold, and molecular properties.

Users define their target (via PDB, mmCIF, or FASTA files) and choose a "modality" (the binder format) and specific properties (like humanization or protease resistance). The system then generates and ranks candidates based on confidence scores and structural fit.

Who it’s for

BC2 is designed for researchers and biologists who need to create custom protein binders for experiments, whether they are designing antibodies (VHH, scFv, Fab), small peptides, or entirely new proteins from scratch.

Highlights

• Diverse Modalities: Supports de novo binders, linear and cyclic peptides, VHH, scFv, Fab, and Ankyrin Repeat proteins (ARP).
• Flexible Targeting: Allows for focused epitopes (hotspots), avoiding off-targets (detargeting), and binding to receptor assemblies or disordered targets.
• Conformational Design: Can design binders that change shape or fold upon binding (induced fit or fold switching).
• Curation Tools: Includes built-in ranking and filtering tools to help users select the best candidates for experimental testing.
• Property Control: Offers flags for humanization, protease stability, and disulfide staples.