ワンクリックでManusで任意のスキルを実行

protein-structure

スター34

フォーク3

更新日2026年7月1日 10:02

Protein structure & protein language models (mirrors Claude Science's protein structure/language-model skill). PDB/UniProt pull & compare structures, Biopython parsing, ESM sequence representation; structure prediction (OpenFold3/Boltz-2/Evo2) via BioNeMo placeholder (needs GPU/NVIDIA). Use for protein structure, PDB, structure comparison, protein language models.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

JWM0203

JWM0203/Claude-Science-for-win

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

protein-structure

Toolchain

Locally reproducible: connectors/pdb.py (structure search/download), connectors/uniprot.py (sequences/annotation), Biopython (parsing/comparison), ESM (if installed, sequence representation).
External access required (placeholder): structure prediction OpenFold3 / Boltz-2 and the Evo 2 sequence model via connectors/_gated.py::bionemo (needs NVIDIA_API_KEY or a local GPU deployment). When unconfigured, report "external access required" — do not return fake structures.

Standard workflow

Pull target structure/sequence from UniProt/PDB (record accession) -> inputs/.
Structure comparison / pocket analysis / sequence representation -> outputs/.
Prediction: run ESM sequence-level locally; structure prediction via bionemo (if configured), otherwise report unavailable.
Figures (structure render / metrics) via figure-maker.

Hard rules

Distinguish experimental (PDB) from predicted structures; note source and confidence (pLDDT). Unavailable hosted models = say external access is required.

このリポジトリの他の Skills

同じリポジトリ

cheminformatics

JWM0203/Claude-Science-for-win

Cheminformatics (mirrors Claude Science's cheminformatics skill). RDKit for descriptors/fingerprints/similarity and molecule drawing, ChEMBL for bioactivity, SAR/screening. Use for small molecules, SMILES, molecular descriptors, ChEMBL, medicinal chemistry. Needs the sci-bio env (rdkit).

2026-07-0134

crispr-screen

JWM0203/Claude-Science-for-win

CRISPR screen design & analysis (mirrors Claude Science's CRISPR screen skill). Designs guide libraries, runs enrichment/depletion analysis on read counts, and traces every hit back to its sgRNA support. Use for sgRNA libraries, screen counts, enrichment analysis.

2026-07-0134

data-analysis

JWM0203/Claude-Science-for-win

Reproducible data analysis. Uses pandas/numpy/scipy/statsmodels for cleaning, statistics, and modeling; scripts and random seeds all land in the run, and every reported number is recomputable. Use to analyze tabular/matrix data, run statistical tests, or model.

2026-07-0134

figure-maker

JWM0203/Claude-Science-for-win

Publication-grade figures. Uses matplotlib to generate figures from a script (never hand-drawn, never pasted from an unknown source); each figure has a rerunnable script. Layer the global CCF-Figure skill for figure conventions / top-venue layout. Use to draw paper figures and iterate to publication quality.

2026-07-0134

literature-review

JWM0203/Claude-Science-for-win

Provenance-backed literature review. Searches PubMed / Europe PMC / arXiv (connectors/literature.py), dedupes, synthesizes by theme, and attaches a DOI/PMID to every conclusion. Use when you need a literature review, evidence for a claim, or a related-work section.

2026-07-0134

manuscript

JWM0203/Claude-Science-for-win

Manuscript writing and iteration to publication quality. Organizes IMRaD structure, inserts run-generated figures/tables, manages references, and produces docx/pdf/md. Layer the global docx / doc-coauthoring / pdf skills. Use to write a paper draft, methods, results, and iterate.

2026-07-0134

name	protein-structure
description	Protein structure & protein language models (mirrors Claude Science's protein structure/language-model skill). PDB/UniProt pull & compare structures, Biopython parsing, ESM sequence representation; structure prediction (OpenFold3/Boltz-2/Evo2) via BioNeMo placeholder (needs GPU/NVIDIA). Use for protein structure, PDB, structure comparison, protein language models.

protein-structure

Toolchain

Locally reproducible: connectors/pdb.py (structure search/download), connectors/uniprot.py (sequences/annotation), Biopython (parsing/comparison), ESM (if installed, sequence representation).
External access required (placeholder): structure prediction OpenFold3 / Boltz-2 and the Evo 2 sequence model via connectors/_gated.py::bionemo (needs NVIDIA_API_KEY or a local GPU deployment). When unconfigured, report "external access required" — do not return fake structures.

Standard workflow

Pull target structure/sequence from UniProt/PDB (record accession) -> inputs/.
Structure comparison / pocket analysis / sequence representation -> outputs/.
Prediction: run ESM sequence-level locally; structure prediction via bionemo (if configured), otherwise report unavailable.
Figures (structure render / metrics) via figure-maker.

Hard rules

Distinguish experimental (PDB) from predicted structures; note source and confidence (pLDDT). Unavailable hosted models = say external access is required.