Protein Mining

自动化蛋白质挖掘完整流程，从文献调研到潜在蛋白确定，生成可发表的学术成果

Security Scan

The skill's high-level purpose (automating protein mining) is plausible, but the SKILL.md and registry metadata are inconsistent about what software exists and what will be run — the skill assumes many heavy bioinformatics tools and a 'protein-mining' CLI that are not provided or declared, which is a notable incoherence the user should resolve before trusting/automating it.

Purpose & Capability

The described pipeline (MSA, HMM, phylogeny, AlphaFold, Foldseek, etc.) matches the stated purpose, but the registry 'requires.bins' lists only mafft, hmmbuild, iqtree while the SKILL.md lists many additional dependencies (CD-HIT, TrimAl, PAML, CAFE, AlphaFold2/ColabFold, Foldseek, PyMOL, etc.). The usage examples also reference a 'protein-mining' CLI that is not provided or declared. This mismatch suggests the skill assumes external software/environment setup that is not documented in metadata.

Instruction Scope

The SKILL.md outlines expected runtime actions (literature search, downloading sequences from UniProt/NCBI, running alignment/HMM/phylogeny and AlphaFold), all of which are relevant to the task. It does not instruct reading unrelated local files or accessing secrets, but it implicitly requires network access to public databases and potentially third-party services (e.g., ColabFold/AlphaFold on remote GPU instances). There are no explicit instructions to transmit data to unknown endpoints, but using Colab/third-party services can result in data leaving the local environment.

Install Mechanism

There is no install spec (instruction-only skill), so nothing is written to disk by the skill itself. That reduces immediate supply-chain risk. However, because the skill expects many external tools and an undocumented 'protein-mining' CLI, the user must manually install/trust these binaries from outside sources.

Credentials

The skill requests no environment variables or credentials, which is appropriate. However, SKILL.md expects use of external services (AlphaFold/ColabFold) that often require accounts or tokens in practice; those are not declared. Also heavy compute/GPU access will be required for structure prediction but this is a resource, not a secret.

Persistence & Privilege

The skill is not always-enabled and is user-invocable; it does not request persistent system privileges or modify other skills. Autonomous invocation is permitted (platform default) but not combined with other high-risk flags.

What to consider before installing

This skill describes a legitimate, heavy bioinformatics pipeline but the metadata is incomplete: it declares only three binaries while the instructions require many more tools (CD-HIT, TrimAl, PAML, CAFE, AlphaFold/ColabFold, Foldseek, PyMOL, Python packages, etc.) and it mentions a 'protein-mining' CLI that is not provided. Before installing or running it: (1) confirm where the 'protein-mining' executable comes from and obtain a trustworthy install method or source code; (2) ensure all required bioinformatics binaries are installed from verified projects (not random URLs); (3) be aware that using AlphaFold/ColabFold or Google Colab may send sequence data to third-party services — avoid submitting sensitive sequences; (4) run the workflow in an isolated environment (container/VM) if you will allow the agent to invoke commands autonomously; (5) ask the author for an explicit install script, dependency list, and provenance for the CLI. If you cannot verify those, treat the skill as untrusted and do not let it run commands on sensitive systems or data.

Protein Mining Skill - 蛋白质挖掘自动化流程

本skill实现了系统化的蛋白质挖掘工作流，包含7个核心模块，每个模块产生可交付的实验结果和可视化图表，适用于高质量学术论文发表。

---

Module 1: Literature Survey & Target Identification

模块1：文献调研与靶标确定

Objectives

• 系统性文献检索与分析
• 确定研究靶标蛋白家族
• 建立研究假设和科学问题

Workflow

1. 文献数据库检索

• PubMed/Google Scholar关键词搜索
• 检索策略：蛋白家族名 + 功能 + 物种
• 时间范围：近10年高影响因子文献优先

2. 文献筛选与分析

• 提取关键蛋白序列信息
• 总结已知功能域和保守位点
• 识别研究空白和创新点

3. 靶标蛋白确定

• 选择代表性种子序列（seed sequences）
• 记录UniProt/NCBI登录号
• 确定研究物种范围

Deliverables

• 01_literature_review.pdf - 文献综述报告
• 01_target_proteins.csv - 靶标蛋白列表（含登录号、物种、功能注释）
• 01_research_hypothesis.md - 研究假设和科学问题陈述

---

Module 2: Protein Sequence Collection

模块2：蛋白序列收集

Objectives

• 从公共数据库获取目标蛋白序列
• 序列质量控制和去冗余
• 构建初始序列数据集

Deliverables

• 02_raw_sequences.fasta - 原始序列集
• 02_filtered_sequences.fasta - 质控后序列集
• 02_nr_sequences.fasta - 非冗余序列集
• 02_sequence_statistics.csv - 序列统计信息
• 02_length_distribution.png - 序列长度分布图

---

Module 3: HMM Profile Construction

模块3：隐马尔可夫模型构建

Objectives

• 构建蛋白家族特异性HMM模型
• 验证模型敏感性和特异性
• 优化模型参数

Deliverables

• 03_msa_alignment.fasta - 多序列比对结果
• 03_protein_family.hmm - HMM模型文件
• 03_model_validation.csv - 模型验证统计
• 03_roc_curve.png - ROC曲线图

---

Module 4: Cross-Species Sequence Search

模块4：跨物种序列搜索

Objectives

• 使用HMM模型进行全基因组搜索
• 跨物种同源蛋白识别
• 构建候选蛋白数据集

Deliverables

• 04_hmm_search_results.tbl - HMM搜索原始结果
• 04_candidate_proteins.fasta - 候选蛋白序列
• 04_species_distribution.csv - 物种分布统计
• 04_species_heatmap.png - 物种分布热图

---

Module 5: Structure-Based Search (Foldseek)

模块5：基于结构的搜索

Objectives

• 利用AlphaFold2预测结构
• Foldseek结构相似性搜索
• 识别远程同源蛋白

Deliverables

• 05_predicted_structures/ - AlphaFold2预测结构
• 05_plddt_scores.csv - 结构置信度评分
• 05_foldseek_results.m8 - Foldseek搜索结果

---

Module 6: Phylogenetic Analysis

模块6：进化树构建与分析

Objectives

• 构建高质量系统发育树
• 进化关系分析
• 功能分化预测

Deliverables

• 06_refined_alignment.fasta - 优化后的多序列比对
• 06_phylogenetic_tree.nwk - Newick格式进化树
• 06_tree_visualization.pdf - 高质量进化树图
• 06_positive_selection_sites.csv - 正选择位点

---

Module 7: Candidate Protein Prioritization

模块7：候选蛋白优先级排序

Objectives

• 整合多维度数据
• 候选蛋白评分与排序
• 确定高优先级实验验证靶标

Deliverables

• 07_feature_matrix.csv - 候选蛋白特征矩阵
• 07_scoring_results.csv - 综合评分结果
• 07_top_candidates.csv - Top 20候选蛋白列表
• 07_experimental_design.md - 实验验证方案建议

---

Technical Requirements

Software Dependencies

• HMMER 3.3+, MAFFT 7.0+, CD-HIT 4.8+, TrimAl 1.4+
• IQ-TREE 2.0+, PAML 4.9+, CAFE 5.0+
• AlphaFold2 / ColabFold, Foldseek 8.0+, PyMOL 2.5+
• Biopython 1.79+, pandas, numpy, scipy, matplotlib, seaborn, scikit-learn

Computational Resources

• CPU: ≥16核心（推荐32核心）
• RAM: ≥64GB（推荐128GB）
• GPU: 用于AlphaFold2结构预测（推荐NVIDIA A100/V100）

---

Usage

# 启动完整流程
protein-mining --target "ABC transporter" --species plants

# 分步执行
protein-mining --module 1 --target "ABC transporter"

---

Citation

• HMMER: Eddy SR. (2011) PLoS Comput Biol.
• IQ-TREE: Nguyen et al. (2015) Mol Biol Evol.
• AlphaFold2: Jumper et al. (2021) Nature.
• Foldseek: van Kempen et al. (2023) Nat Biotechnol.