bindcraft
Guide CUDA-enabled BindCraft protein-binder design from target PDB preparation through AF2/MPNN/PyRosetta execution, filtering, output analysis, and conservative troubleshooting.
- License 许可
- MIT
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- bindcraft
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Guide CUDA-enabled BindCraft protein-binder design from target PDB preparation through AF2/MPNN/PyRosetta execution, filtering, output analysis, and conservative troubleshooting.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/configuration.md3.2 KB
- 参考文档references/installation.md3.3 KB
- 参考文档references/repo-provenance.md1.5 KB
- 参考文档references/repo-routing-metadata.json328 B
- 参考文档references/troubleshooting.md4.0 KB
03 SKILL.md 正文
BindCraft
BindCraft is a GPU-first de-novo protein-binder design pipeline that combines AlphaFold2 backpropagation, ProteinMPNN sequence redesign, AF2 complex/monomer validation, and PyRosetta relaxation and interface scoring. Use this skill to plan and operate a reproducible campaign; do not treat it as a generic CPU protein-design library.
Route the request
- Prepare or repair a target PDB and target JSON: read target-preparation. It covers chain selection, hotspot syntax, binder lengths, and safe validation.
- Install prerequisites, choose presets, build a launch, run or resume a campaign: read design-pipeline. It owns CUDA/JAX, AF2-weight, MPNN, PyRosetta, direct, and SLURM decisions.
- Inspect results, rejection causes, scores, and ranked binders: read results-analysis. It owns output reconciliation and conservative metric interpretation.
Before using any route, read installation for external prerequisites and licensing, configuration for the three JSON families, and troubleshooting for cross-cutting failures. Read repo-provenance when checking whether this graph matches a repository revision.
Minimal operating contract
- Work on Linux with Python 3.10 and an NVIDIA GPU. The main design path requires a CUDA-visible JAX runtime; a CPU import is not a valid substitute.
- Obtain the AlphaFold2 parameter bundle, configure its directory, and ensure DSSP and DAlphaBall are readable/executable. BindCraft's installer downloads large external artifacts; review and run setup commands manually rather than blindly executing a bundled installer.
- Install ColabDesign and PyRosetta in a private, compatible environment. PyRosetta may have commercial-license restrictions; resolve those before a production run.
- Validate the target JSON and PDB, then use the design route's command builder to print a direct or SLURM command. The builder never executes or submits.
- Use a distinct writable
design_pathper campaign. Expect hundreds or thousands of target-dependent trajectories for difficult targets; monitor GPU memory, disk use,failure_csv.csv, and acceptance rate. - Rank and select candidates from recorded artifacts, not from confidence
metrics alone. BindCraft documents
Average_i_pTMas a useful binding binary/ranking signal, not an affinity measurement.
Safe helpers
- From the generated skill root,
python scripts/check_bindcraft_env.pyreports import/backend/asset readiness and never installs, downloads, or launches a design. - From the generated skill root,
python scripts/validate_bindcraft_config.pychecks the target, filter, and advanced JSON contracts without editing them. - The focused routes link additional read-only target and results helpers. All paths in examples are placeholders that must be replaced on the launch host.
Non-goals and stop conditions
This graph does not download AF2 weights, submit SLURM jobs, run a full design campaign, promise a binder, or infer experimental affinity. Stop and repair the specific prerequisite when CUDA/JAX, ColabDesign, PyRosetta, AF2 weights, DSSP, DAlphaBall, PDB chains, settings, output permissions, or disk/VRAM capacity are not verified. A successful JSON/PDB check or generated command is not evidence that the GPU design loop will complete.
04 同仓库更多 Skills
ci-cd-and-automation
Automates CI/CD pipeline setup. Use when setting up or modifying build and deployment pipelines. Use when you need to automate quality gates, configure test runners in CI, or establish deployment strategies.
chrome-devtools
Uses Chrome DevTools via MCP for efficient debugging, troubleshooting and browser automation. Use when debugging web pages, automating browser interactions, analyzing performance, or inspecting network requests. This skill does not apply to `--slim` mode (MCP configuration).
chrome-cdp
Drive a real Chrome via the Chrome DevTools Protocol (CDP). Ensures a debuggable Chrome is running (launching it if needed) so the agent can navigate, read page content, inspect the DOM, run JavaScript, and take screenshots — without the user typing any launch command. Includes a first-class "review a live running app" flow: point Chrome at the app under development and inspect/screenshot the REAL page (what a static-artifact tool cannot do). Backed by Glimpse and the chrome-devtools MCP server. Use when asked to "use chrome", "open chrome", "drive my browser", "read the page in chrome", "review the running app", "screenshot the app", or "/chrome-cdp".
check-identity-pack
Run an AFP 100-point or AUSTRAC safe-harbour identity check over a set of documents, and report exactly what's missing. Use when the user asks to check identity documents, verify someone's ID for onboarding, or assess whether a document pack satisfies Australian identity requirements.
changelog
Auto-generates a changelog from git commits, sprint data, and design documents. Produces both internal and player-facing versions.
cellxgene-census
Query the CZ CELLxGENE Census programmatically for versioned public single-cell and spatial transcriptomics data. Use when you need population-scale cell metadata, gene expression slices, Census summary counts, source H5AD URIs/downloads, embeddings, spatial Census data, or reference atlas comparisons across organisms, tissues, diseases, assays, and cell types. For analyzing your own local single-cell data use scanpy, anndata, or scvi-tools.