medgeclaw-guide
安装方式
先用 Prompt 核对来源,也可切换到安装命令或下载可用的安装包。
用 AI 助手安装
复制提示词,让助手先核对固定版本和依赖。
查看 Prompt收起 Prompt
先检查当前 Agent 是否已安装 SkillPI Skill Finder。如果尚未安装,请按官方仓库的说明安装并启用:https://github.com/Findskills/skillpi。然后使用 Skill Finder 打开此 SkillPI 详情页并定位 "medgeclaw-guide":https://skillpi.com/zh/skills/deepskill-9fbedd7df9722ae698221dafd699fde795b4e16f523583347b626bba7f5814df。安装前核对上游来源和版本:https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/domains/biomedical/medgeclaw-guide/SKILL.md。来源不一致时停止并说明差异。核对通过后,用 Skill Finder 安装目标 Skill,并报告安装路径和来源版本。
此命令从 GitHub 来源仓库获取指定技能。运行前请先查看源文件。
npx skills add https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/tree/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/domains/biomedical/medgeclaw-guide --skill medgeclaw-guide来源与配置
查看 GitHub 源码brycewang-stanford/Auto-Empirical-Research-Skills固定提交 bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba
配置方式待核实
请查看原始来源中的适用工具、依赖和权限要求。
安装实测待补充
本条目面向文献与学术工具场景,提供“medgeclaw-guide”技能的目录入口。公开标签:learning、RNA-seq、药物发现、生物医学。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。
文件浏览
---
name: medgeclaw-guide
description: "AI research assistant for biomedicine, RNA-seq, and drug discovery"
metadata:
openclaw:
emoji: "💊"
category: "domains"
subcategory: "biomedical"
keywords: ["biomedicine", "RNA-seq", "drug discovery", "clinical AI", "medical NLP", "bioinformatics"]
source: "wentor-research-plugins"
---
# MedgeClaw Guide
## Overview
MedgeClaw is a conceptual framework for AI-powered biomedical research assistance, integrating natural language processing for medical literature, computational biology pipelines, and drug discovery workflows. The name reflects the integration of Medical knowledge Edge (cutting-edge biomedical AI) with the Claw agent pattern for autonomous research execution.
Biomedical research is uniquely suited for AI augmentation because it generates massive, heterogeneous data -- genomic sequences, clinical records, imaging data, molecular structures, and published literature -- that exceeds the capacity of individual researchers to synthesize. AI systems that can navigate across these data types, identify patterns, and suggest hypotheses accelerate the pace of discovery.
This guide covers the key computational methods in biomedical AI research: medical NLP for literature mining, RNA-seq analysis pipelines, drug discovery computational workflows, and the integration patterns that connect these components into coherent research workflows. The focus is on methods that are reproducible, validated, and suitable for publication in biomedical journals.
## Medical NLP and Literature Mining
### Biomedical Named Entity Recognition
```python
# Biomedical NER using scispaCy
import scispacy
import spacy
from scispacy.linking import EntityLinker
# Load biomedical NER model
nlp = spacy.load("en_ner_bionlp13cg_md")
# Add UMLS entity linker for concept normalization
nlp.add_pipe("scispacy_linker", config={
"resolve_abbreviations": True,
"linker_name": "umls",
})
def extract_biomedical_entities(text: str) -> dict:
"""
Extract and normalize biomedical entities from text.
Returns genes, chemicals, diseases, and their UMLS mappings.
"""
doc = nlp(text)
entities = {
"genes": [],
"chemicals": [],
"diseases": [],
"other": [],
}
category_map = {
"GENE_OR_GENE_PRODUCT": "genes",
"SIMPLE_CHEMICAL": "chemicals",
"CANCER": "diseases",
"ORGAN": "other",
"CELL": "other",
}
for ent in doc.ents:
category = category_map.get(ent.label_, "other")
entity_info = {
"text": ent.text,
"label": ent.label_,
"start": ent.start_char,
"end": ent.end_char,
}
# Add UMLS links if available
if hasattr(ent, "_") and hasattr(ent._, "kb_ents"):
if ent._.kb_ents:
top_link = ent._.kb_ents[0]
entity_info["umls_cui"] = top_link[0]
entity_info["confidence"] = round(top_link[1], 3)
entities[category].append(entity_info)
return entities
```
### Systematic Literature Search Pipeline
```python
from Bio import Entrez
import time
Entrez.email = "researcher@university.edu"
def systematic_pubmed_search(
query: str,
max_results: int = 1000,
date_range: tuple = ("2020/01/01", "2025/12/31"),
) -> list:
"""
Conduct a systematic PubMed search with structured result extraction.
Suitable for systematic reviews and meta-analyses.
"""
# Step 1: Search PubMed
handle = Entrez.esearch(
db="pubmed",
term=query,
retmax=max_results,
datetype="pdat",
mindate=date_range[0],
maxdate=date_range[1],
sort="relevance",
)
results = Entrez.read(handle)
handle.close()
pmids = results["IdList"]
print(f"Found {results['Count']} results, retrieving {len(pmids)}")
# Step 2: Fetch article details in batches
articles = []
batch_size = 100
for i in range(0, len(pmids), batch_size):
batch = pmids[i:i + batch_size]
handle = Entrez.efetch(
db="pubmed", id=",".join(batch),
rettype="xml", retmode="xml"
)
records = Entrez.read(handle)
handle.close()
for article in records["PubmedArticle"]:
medline = article["MedlineCitation"]
art = medline["Article"]
articles.append({
"pmid": str(medline["PMID"]),
"title": art["ArticleTitle"],
"abstract": art.get("Abstract", {}).get("AbstractText", [""])[0],
"journal": art["Journal"]["Title"],
"year": art["Journal"]["JournalIssue"]["PubDate"].get("Year", "N/A"),
"mesh_terms": [
d["DescriptorName"]
for d in medline.get("MeshHeadingList", [])
] if "MeshHeadingList" in medline else [],
})
time.sleep(0.4) # Respect NCBI rate limits
return articles
```
## RNA-seq Analysis
### Complete DESeq2 Workflow
```r
# Complete RNA-seq differential expression analysis with DESeq2
# This is the standard workflow for biomedical RNA-seq papers
library(DESeq2)
library(ggplot2)
library(EnhancedVolcano)
library(clusterProfiler)
library(org.Hs.eg.db)
# --- 1. Load count matrix and metadata ---
counts <- read.csv("raw_counts.csv", row.names = 1)
coldata <- read.csv("sample_info.csv", row.names = 1)
# Verify sample order matches
stopifnot(all(colnames(counts) == rownames(coldata)))
# --- 2. Create DESeq2 object ---
dds <- DESeqDataSetFromMatrix(
countData = counts,
colData = coldata,
design = ~ condition # Simple two-group comparison
)
# Pre-filtering: remove low-count genes
keep <- rowSums(counts(dds)) >= 10
dds <- dds[keep, ]
# --- 3. Run differential expression ---
dds <- DESeq(dds)
res <- results(dds, contrast = c("condition", "treatment", "control"),
alpha = 0.05)
# Summary
summary(res)
# --- 4. Results with shrinkage (recommended for visualization) ---
res_shrunk <- lfcShrink(dds, coef = "condition_treatment_vs_control",
type = "apeglm")
# --- 5. Export significant genes ---
sig_genes <- subset(res, padj < 0.05 & abs(log2FoldChange) > 1)
write.csv(as.data.frame(sig_genes), "significant_genes.csv")
```
### Quality Control Metrics
| Metric | Expected Range | Concern If |
|--------|---------------|------------|
| Total reads | 20-50M per sample | < 10M |
| Mapping rate | > 80% | < 70% |
| rRNA contamination | < 5% | > 10% |
| GC content | ~42% (human) | Bimodal distribution |
| Duplication rate | < 30% (mRNA) | > 50% |
| Gene body coverage | Uniform 5' to 3' | Strong 3' bias |
| PCA | Samples cluster by condition | Outlier samples |
## Drug Discovery Computational Methods
### Virtual Screening Pipeline
```python
# Molecular docking workflow using RDKit and AutoDock Vina
from rdkit import Chem
from rdkit.Chem import AllChem, Descriptors, Lipinski
import subprocess
def prepare_ligands(smiles_list: list) -> list:
"""
Prepare ligands for virtual screening.
Apply Lipinski's Rule of Five and generate 3D conformers.
"""
prepared = []
for smiles in smiles_list:
mol = Chem.MolFromSmiles(smiles)
if mol is None:
continue
# Lipinski's Rule of Five filter
mw = Descriptors.MolWt(mol)
logp = Descriptors.MolLogP(mol)
hbd = Descriptors.NumHDonors(mol)
hba = Descriptors.NumHAcceptors(mol)
if mw > 500 or logp > 5 or hbd > 5 or hba > 10:
continue # Fails Ro5
# Generate 3D conformer
mol_h = Chem.AddHs(mol)
AllChem.EmbedMolecule(mol_h, AllChem.ETKDG())
AllChem.MMFFOptimizeMolecule(mol_h)
prepared.append({
"smiles": smiles,
"mol": mol_h,
"mw": round(mw, 2),
"logp": round(logp, 2),
"hbd": hbd,
"hba": hba,
})
return prepared
def compute_admet_properties(mol) -> dict:
"""Compute ADMET-relevant molecular descriptors."""
return {
"tpsa": round(Descriptors.TPSA(mol), 2), # Topological polar surface area
"rotatable_bonds": Descriptors.NumRotatableBonds(mol),
"aromatic_rings": Descriptors.NumAromaticRings(mol),
"fraction_csp3": round(Descriptors.FractionCSP3(mol), 3), # Drug-likeness
"qed": round(Descriptors.qed(mol), 3), # Quantitative drug-likeness
}
```
### Target-Disease Association Analysis
```python
def query_open_targets(target_id: str, disease_id: str) -> dict:
"""
Query Open Targets Platform for target-disease association evidence.
"""
import requests
query = """
query targetDiseaseAssociation($target: String!, $disease: String!) {
disease(efoId: $disease) {
name
associatedTargets(Bs: [$target]) {
rows {
target { approvedSymbol }
score
datatypeScores {
componentId: id
score
}
}
}
}
}
"""
response = requests.post(
"https://api.platform.opentargets.org/api/v4/graphql",
json={"query": query, "variables": {"target": target_id, "disease": disease_id}},
)
return response.json()
```
## Clinical AI Applications
### Clinical NLP Patterns
```
Common clinical NLP tasks for research:
1. CLINICAL TEXT DE-IDENTIFICATION
- Remove PHI (Protected Health Information)
- Tools: Philter, NLM Scrubber, custom regex + NER
- Validation: Must achieve >95% recall for PHI
2. CLINICAL CODING
- Assign ICD-10, CPT, SNOMED-CT codes to clinical notes
- Approaches: Rule-based, ML classification, LLM extraction
- Evaluation: Precision/recall per code family
3. RELATION EXTRACTION
- Drug-disease, drug-adverse event, gene-disease relationships
- From clinical notes, discharge summaries, pathology reports
- Output: Knowledge graphs for downstream analysis
4. TEMPORAL INFORMATION EXTRACTION
- Disease onset, treatment timeline, outcome timing
- Critical for longitudinal studies and survival analysis
- Tools: SUTime, HeidelTime, custom models
```
## Best Practices
- **Validate AI predictions experimentally.** Computational predictions are hypotheses until confirmed in the lab.
- **Use standard file formats.** FASTQ for sequencing, SDF/MOL2 for molecules, FASTA for sequences, VCF for variants.
- **Follow FAIR data principles.** Findable, Accessible, Interoperable, Reusable data management.
- **De-identify clinical data before any AI processing.** HIPAA and GDPR compliance is non-negotiable.
- **Report computational methods in full detail.** Software versions, parameters, random seeds, and hardware specs.
- **Pre-register clinical AI studies.** Use SPIRIT-AI or CONSORT-AI reporting guidelines.
## References
- [DESeq2](https://bioconductor.org/packages/DESeq2/) -- Standard RNA-seq differential expression tool
- [scispaCy](https://allenai.github.io/scispacy/) -- Biomedical NLP models for spaCy
- [Open Targets Platform](https://platform.opentargets.org/) -- Target-disease association evidence
- [RDKit](https://www.rdkit.org/) -- Cheminformatics toolkit
- Love, M. I., Huber, W., & Anders, S. (2014). Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology, 15, 550.
版本历史
-
v2026.07
GitHub ↗
原版更新说明
## Highlights - **First tagged release.** Every gate is green: `make check` (validate + 183 unit tests + eval-harness + numeric benchmarks), validate-catalog, quality-evals, and OpenSSF Scorecard on `main`. - **Rigor double loop at 13 method families.** Each family closes taxonomy tag → eval scenario (does the agent avoid the method's known traps?) → deterministic numeric benchmark (does it recover the true effect?). New this cycle: heterogeneous treatment effects (CATE) and quantile treatment effects (QTE). - **Paper-WorkFlow flagship at 29/29 executable rigor gates**, now merged to its `main` with a drift-gated README badge — the credibility artifact is checked by CI, so it can never silently go stale ([RIGOR.md](https://github.com/brycewang-stanford/Paper-WorkFlow/blob/main/RIGOR.md)). - **1,150 vendored skills / 69 collections**, all license-audited, security-scanned, tracked in a machine-readable catalog with CI-enforced freshness, and importable as a single whole-repo router skill. - **Hardened weekly upstream syncs**: synced trees must pass the full gate before merge; this cycle both sync PRs were verified, patched (StatsPAI's SkillOpt execution-gate card restored), and merged. - **First-party plugin marketplace** (Python / R / Stata empirical-analysis plugins + guide) shipped in the previous cycle, now under a tagged version. ## Known Follow-Ups - Ecosystem benchmark comparison (AERS vs econometrics agents on the shared benchmark suite) — [PLAN-2026-07 week 3](https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/main/docs/PLAN-2026-07.md). - Decomposition (Oaxaca–Blinder) and shift-share (Bartik) method families (weeks 3–4). - GitHub Pages catalog site, README demo screenshots, five-language README consistency script (week 4). ## Catalog - Top-level collections: **69** - `SKILL.md` files: **1150** - Cataloged agent tools: **335** ([`tools/CATALOG.md`](../tools/CATALOG.md)) ## Methodological rigor - Method families in the coverage map: **13** ([`RIGOR_COVERAGE.md`](RIGOR_COVERAGE.md)) - Eval scenarios: **26** (138 rubric items, 122 auto-checkable) ([`../eval-harness/`](../eval-harness/README.md)) - Benchmark tasks: **13** (70 gold checks, 58 required) ([`../benchmark/`](../benchmark/README.md)) ### Method families (skills tagged) - `iv`: 27 - `rdd`: 25 - `panel-fe`: 19 - `staggered-did`: 17 - `bayesian`: 13 - `did`: 10 - `event-study`: 10 - `matching`: 10 - `synthetic-control`: 10 - `cate`: 8 - `dml`: 8 - `survival`: 3 - `quantile`: 2 ### Eval scenarios by category - `causal-identification`: 14 - `citation-hygiene`: 1 - `reproducibility`: 2 - `research-integrity`: 1 - `runtime-safety`: 1 - `statistical-validity`: 2 - `writing-compliance`: 3 - `writing-style`: 2 ## License buckets - AGPL-3.0: 1 - Apache-2.0: 1 - CC-BY-NC-4.0 (non-commercial): 1 - CC-BY-SA-4.0: 1 - CC-BY-SA-4.0 (repository default): 4 - GPL-3.0: 1 - MIT: 35 - MIT Non-Commercial: 1 - Mixed (per-skill; see folder): 2 - UNKNOWN - check upstream: 22 ## Quality gates - `make check` — catalog/provenance/audit/eval freshness, repo hygiene, Paper-WorkFlow gates, stdlib unit tests, eval-harness lint + graded smoke, and the strict numeric benchmark (reference-gold drift fails the build). - External links and OpenSSF Scorecard run on their scheduled workflows.
当前收录版本
0.0.1
质量评测
满分 100。前四项来自 DeepSkill 收录的质量报告。
评分依据
提供生物医学、RNA-seq 和药物发现领域的完整 AI 研究助手功能
评分依据
description 明确描述领域和核心能力
评分依据
文档结构清晰,分多个专业章节组织内容
评分依据
代码示例来源明确,使用 scispaCy、DESeq2、RDKit 等标准工具
计算方式
取前四项原始分值的算术平均,四舍五入取整。可通过收录报告核对各项评分与说明,运行记录需单独查证。
权限与使用前检查
根据当前主文档自动整理,命中内容可能是示例或防护说明;不是实际权限声明,也未覆盖附件、外部依赖或运行行为。
未从主文档识别到上述常见操作,不能据此判断无需权限或没有风险。
文档引用的域名或地址
这些地址可能用于文档链接或示例,不是已批准的联网清单。使用前核对实际目的地及发送的数据。
- allenai.github.io
- api.platform.opentargets.org
- bioconductor.org
- platform.opentargets.org
- www.rdkit.org
安全评级与访问范围
risk · 静态证据核验,不代表运行安全保证。
使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地;本站的核验不会授予任何运行权限。
用户讨论
还没有已核实的点评。