Automate de novo protein binder design with AF2 and PyRosetta validation.
npx skills add https://github.com/LiorZ/protein-design-skills --skill bindcraft-liorz
Please help me install this Agent Skill. Skill: bindcraft Source: https://github.com/LiorZ/protein-design-skills/tree/main/skills/bindcraft Command: npx skills add https://github.com/LiorZ/protein-design-skills --skill bindcraft-liorz
BindCraft addresses the challenge of efficiently designing de novo protein binders by providing an automated workflow that couples AF2 backpropagation, ProteinMPNN redesign, AF2 reprediction, and a PyRosetta-based filter, enabling end-to-end campaigns from target preparation to ranked designs.
End-to-end binder design workflow that samples trajectories, hallucination via AF2, MPNN redesign, AF2 reprediction, interface scoring, and a multi-filter selection process.
Supports multiple flavors (default_4stage_multimer, betasheet_4stage_multimer, peptide_3stage_multimer) and modifiers (_mpnn, _flexible, _hardtarget) for diverse topologies.
Outputs structured campaigns with Trajectory, MPNN, Accepted/Ranked, and Rejected directories plus CSV metrics for triage.
Use cases include mini-protein binders, beta-sheet binders, and short peptides against protein targets; it supports SLURM campaigns and independent cross-validation.
Run BindCraft with your target JSON, filter preset, and advanced settings to start a binder-design campaign.
High-intent search queries and answers about installing and using this skill.
De novo protein binder design is automated by hallucinating binders against a target via AF2, redesigning sequences with ProteinMPNN, and validating with AF2 and PyRosetta. This pipeline processes everything from target trimming to ranked designs.
Hallucination-based binder design works by using AF2 to hallucinate and sample protein trajectories, redesigning sequences with ProteinMPNN, and repredicting structures with AF2. PyRosetta then applies multi-filter scoring to validate the interface.
Yes, you can design short peptide and beta-sheet binders using this hallucination workflow. It supports specific flavors including peptide_3stage_multimer for short peptides and betasheet_4stage_multimer for beta-sheet binders against protein targets.
You can run binder design campaigns on a SLURM cluster by configuring the pipeline for distributed execution. The workflow supports SLURM-based runs, generating structured Trajectory, MPNN, Accepted, and Rejected directories with detailed CSV metrics for triage.
Yes, you need to prepare specific JSON files before starting a binder design campaign. The pipeline enforces a three-JSON contract requiring settings_target, settings_filters, and settings_advanced files to configure your AF2 and PyRosetta validation workflow.
Validation limitations depend on chosen filter presets and target difficulty. The pipeline offers modifiers like _hardtarget for difficult cases, but success is constrained by AF2 reprediction accuracy and the strictness of the PyRosetta multi-filter selection process.
zongtingweiPredict protein complex structures and assess interface confidence with AlphaFold2-Multimer.
alex-hhCross-validate 15-PGDH binder designs using Boltz-2 structure prediction.
adaptyvbioAutomate end-to-end binder design with BindCraft and AF2 validation.
junior1pGuide protein binder design workflows with BoltzGen, BindCraft, and Chai validation.
