How to Use BindCraft

BindCraft is a de novo protein binder design pipeline built for one-shot generation of functional binders against protein targets. The Nature paper describes a workflow that uses structure prediction as the core design engine, allowing high-affinity binders to emerge from iterative sequence-structure optimization rather than from very large experimental screening libraries.

On Neurosnap, researchers begin with a Target Structure and can optionally restrict the campaign to specific Target Chains or Hotspots. Binder size bounds and protocol choices then shape the search, making the service useful for epitope-focused therapeutic design, receptor targeting, and rapid exploration of surfaces that may not already have known binders.

BindCraft repurposes AlphaFold-based complex prediction as a design objective. Candidate binders are generated and repeatedly self-evaluated against the target so that interface packing, predicted confidence, and binding geometry all influence which sequences survive. The published pipeline couples this structural feedback with sequence design and filtering, which is why it moves closer to a one-design, one-binder workflow than older binder-design strategies.

On Neurosnap, Minimum Binder Size, Maximum Binder Size, Design Protocol, Prediction Protocol, Interface Protocol, Template Protocol, and Filters expose the main levers researchers tune when balancing exploration against stringency. Leaving Hotspots blank allows broader site discovery, while explicit hotspots focus the run on a chosen epitope or interaction patch.

Results are best interpreted as a ranked binder campaign rather than a single answer. Researchers compare predicted complex geometry, interface-focused confidence, contact patterns, and quality trends across candidates to decide which binders merit expression, affinity testing, or follow-up maturation.

• Study-fit inputs: BindCraft starts from a target structure and can focus on selected chains or hotspots, which matches many epitope-driven binder campaigns.
• One-shot structural design: The method uses structure prediction inside the design loop, allowing interface quality and foldability to shape candidate generation from the outset.
• Site-directed control: Size bounds and protocol settings let researchers bias the search toward compact binders, peptide-like solutions, beta-rich architectures, or more permissive filtering.
• Candidate triage: Ranked complexes, interface confidence, and design-quality trends make it easier to choose a short list for expression and experimental validation.