@nahisaho/satori 0.13.0 → 0.14.0

package/src/.github/skills/scientific-public-health-data/SKILL.md

@@ -0,0 +1,322 @@
+---
+name: scientific-public-health-data
+description: |
+  公衆衛生データアクセススキル。NHANES 疫学調査データ、MedlinePlus 一般向け
+  健康情報、RxNorm 薬剤標準語彙、ODPHP 健康目標・ガイドライン、
+  Health Disparities 健康格差データ統合パイプライン。
+---
+
+# Scientific Public Health Data
+
+NHANES / MedlinePlus / RxNorm / ODPHP / Health Disparities /
+Guidelines を統合した公衆衛生データアクセスパイプラインを提供する。
+
+## When to Use
+
+- NHANES 疫学調査データ (検査値・アンケート) を取得するとき
+- MedlinePlus で一般向け健康情報を検索するとき
+- RxNorm で薬剤名の標準化・マッピングを行うとき
+- ODPHP Healthy People 目標や健康ガイドラインを参照するとき
+- 健康格差 (Health Disparities) データを分析するとき
+- 臨床ガイドライン (USPSTF/WHO) を検索するとき
+
+---
+
+## Quick Start
+
+## 1. NHANES 疫学調査データ取得
+
+```python
+import requests
+import pandas as pd
+import io
+
+NHANES_BASE = "https://wwwn.cdc.gov/nchs/nhanes"
+
+
+def get_nhanes_dataset(cycle, dataset_name):
+    """
+    NHANES データセット (XPT/SAS 形式) 取得。
+
+    Parameters:
+        cycle: str — 調査サイクル (e.g., "2017-2018", "2019-2020")
+        dataset_name: str — データセット名 (e.g., "DEMO_J", "BIOPRO_J")
+
+    ToolUniverse:
+        NHANES_get_dataset(cycle=cycle, dataset=dataset_name)
+        NHANES_list_datasets(cycle=cycle)
+    """
+    cycle_code = cycle.replace("-", "_")
+    url = f"{NHANES_BASE}/search/DataPage.aspx"
+
+    # XPT ファイルの直接ダウンロード
+    xpt_url = f"https://wwwn.cdc.gov/Nchs/Nhanes/{cycle}/{dataset_name}.XPT"
+    resp = requests.get(xpt_url)
+    resp.raise_for_status()
+
+    df = pd.read_sas(io.BytesIO(resp.content), format="xport")
+    print(f"NHANES {cycle} {dataset_name}: {df.shape[0]} rows × {df.shape[1]} columns")
+    return df
+
+
+def search_nhanes_variables(keyword):
+    """
+    NHANES 変数検索。
+
+    Parameters:
+        keyword: str — 変数名/説明の検索語
+
+    ToolUniverse:
+        NHANES_search_variables(keyword=keyword)
+    """
+    url = f"{NHANES_BASE}/search/variablelist.aspx"
+    params = {"SearchTarget": keyword}
+    resp = requests.get(url, params=params)
+    resp.raise_for_status()
+
+    print(f"NHANES variable search '{keyword}': response received")
+    return resp.text
+```
+
+## 2. MedlinePlus 健康情報検索
+
+```python
+MEDLINEPLUS_API = "https://connect.medlineplus.gov/service"
+MEDLINEPLUS_WS = "https://wsearch.nlm.nih.gov/ws/query"
+
+
+def search_medlineplus_health_topics(query, language="English"):
+    """
+    MedlinePlus 健康トピック検索。
+
+    ToolUniverse:
+        MedlinePlus_search_health_topics(query=query)
+        MedlinePlus_get_health_topic(topic_id=topic_id)
+        MedlinePlus_search_drugs(query=query)
+        MedlinePlus_search_labs(query=query)
+        MedlinePlus_connect(code=code, code_system=system)
+    """
+    params = {
+        "db": "healthTopics",
+        "term": query,
+    }
+    resp = requests.get(MEDLINEPLUS_WS, params=params)
+    resp.raise_for_status()
+
+    # XML response parsing
+    import xml.etree.ElementTree as ET
+    root = ET.fromstring(resp.text)
+
+    results = []
+    for doc in root.findall(".//document"):
+        results.append({
+            "title": doc.find(".//content[@name='title']").text
+            if doc.find(".//content[@name='title']") is not None else "",
+            "url": doc.get("url", ""),
+            "summary": doc.find(".//content[@name='FullSummary']").text[:300]
+            if doc.find(".//content[@name='FullSummary']") is not None else "",
+            "rank": doc.get("rank", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"MedlinePlus search '{query}': {len(df)} health topics")
+    return df
+```
+
+## 3. RxNorm 薬剤標準語彙
+
+```python
+RXNORM_API = "https://rxnav.nlm.nih.gov/REST"
+
+
+def rxnorm_lookup(drug_name):
+    """
+    RxNorm 薬剤名正規化・コードマッピング。
+
+    Parameters:
+        drug_name: str — 薬剤名 (商品名 or 一般名)
+
+    ToolUniverse:
+        RxNorm_get_rxcui(name=drug_name)
+    """
+    resp = requests.get(
+        f"{RXNORM_API}/rxcui.json",
+        params={"name": drug_name}
+    )
+    resp.raise_for_status()
+    data = resp.json()
+
+    rxcui = data.get("idGroup", {}).get("rxnormId", [None])[0]
+    if not rxcui:
+        print(f"RxNorm: '{drug_name}' not found")
+        return None
+
+    # Get properties
+    props_resp = requests.get(f"{RXNORM_API}/rxcui/{rxcui}/properties.json")
+    props_resp.raise_for_status()
+    props = props_resp.json().get("properties", {})
+
+    # Get related concepts
+    related_resp = requests.get(
+        f"{RXNORM_API}/rxcui/{rxcui}/related.json",
+        params={"tty": "IN+BN+SBD+SCD"}
+    )
+    related_resp.raise_for_status()
+    related = related_resp.json()
+
+    result = {
+        "rxcui": rxcui,
+        "name": props.get("name", ""),
+        "tty": props.get("tty", ""),
+        "synonym": props.get("synonym", ""),
+        "related_concepts": [
+            {
+                "rxcui": c.get("rxcui"),
+                "name": c.get("name"),
+                "tty": c.get("tty"),
+            }
+            for group in related.get("relatedGroup", {}).get("conceptGroup", [])
+            for c in group.get("conceptProperties", [])
+        ],
+    }
+    print(f"RxNorm '{drug_name}': RXCUI={rxcui}, TTY={result['tty']}")
+    return result
+```
+
+## 4. Health Disparities データ取得
+
+```python
+HD_API = "https://data.cdc.gov/resource"
+
+
+def get_health_disparities(indicator, dataset_id="pqnx-3xr5"):
+    """
+    CDC 健康格差データ取得。
+
+    Parameters:
+        indicator: str — 健康指標名
+        dataset_id: str — CDC Socrata データセット ID
+
+    ToolUniverse:
+        HealthDisparities_search(query=indicator)
+        HealthDisparities_get_indicators(category=category)
+    """
+    params = {
+        "$where": f"indicator LIKE '%{indicator}%'",
+        "$limit": 1000,
+    }
+    resp = requests.get(f"{HD_API}/{dataset_id}.json", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    df = pd.DataFrame(data)
+    print(f"Health Disparities '{indicator}': {len(df)} records")
+    return df
+```
+
+## 5. ODPHP 健康ガイドライン
+
+```python
+ODPHP_API = "https://health.gov/myhealthfinder/api/v3"
+
+
+def search_health_guidelines(keyword, category=None):
+    """
+    ODPHP MyHealthfinder ガイドライン検索。
+
+    ToolUniverse:
+        ODPHP_search_topics(keyword=keyword)
+        ODPHP_get_topic(topic_id=topic_id)
+    """
+    params = {"keyword": keyword}
+    if category:
+        params["categoryId"] = category
+    resp = requests.get(f"{ODPHP_API}/topicsearch.json", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for topic in data.get("Result", {}).get("Resources", {}).get("Resource", []):
+        results.append({
+            "title": topic.get("Title", ""),
+            "categories": topic.get("Categories", ""),
+            "url": topic.get("AccessibleVersion", ""),
+            "sections": [
+                s.get("Title", "") for s in topic.get("Sections", {}).get("section", [])
+            ],
+        })
+
+    df = pd.DataFrame(results)
+    print(f"ODPHP search '{keyword}': {len(df)} guidelines")
+    return df
+```
+
+## 6. 臨床ガイドライン検索 (USPSTF)
+
+```python
+def search_clinical_guidelines(query, source="uspstf"):
+    """
+    USPSTF/WHO 臨床ガイドライン検索。
+
+    ToolUniverse:
+        Guidelines_search(query=query, source=source)
+        Guidelines_get_recommendations(topic_id=topic_id)
+    """
+    sources = {
+        "uspstf": "https://www.uspreventiveservicestaskforce.org/uspstf/api",
+        "who": "https://app.magicapp.org/api",
+    }
+    base_url = sources.get(source, sources["uspstf"])
+
+    resp = requests.get(f"{base_url}/search", params={"q": query})
+    if resp.status_code == 200:
+        data = resp.json()
+        results = []
+        for item in data.get("results", []):
+            results.append({
+                "title": item.get("title", ""),
+                "grade": item.get("grade", ""),
+                "population": item.get("population", ""),
+                "date": item.get("date", ""),
+                "recommendation": item.get("recommendation", ""),
+            })
+        df = pd.DataFrame(results)
+    else:
+        df = pd.DataFrame()
+
+    print(f"Guidelines ({source}) search '{query}': {len(df)} recommendations")
+    return df
+```
+
+---
+
+## 利用可能ツール
+
+| ToolUniverse カテゴリ | 主なツール |
+|---|---|
+| `nhanes` | `NHANES_get_dataset`, `NHANES_list_datasets`, `NHANES_search_variables` |
+| `health_disparities` | `HealthDisparities_search`, `HealthDisparities_get_indicators` |
+| `medlineplus` | `MedlinePlus_search_health_topics`, `MedlinePlus_get_health_topic`, `MedlinePlus_search_drugs`, `MedlinePlus_search_labs`, `MedlinePlus_connect` |
+| `odphp` | `ODPHP_search_topics`, `ODPHP_get_topic` |
+| `rxnorm` | `RxNorm_get_rxcui` |
+| `guidelines_tools` | `Guidelines_search`, `Guidelines_get_recommendations` |
+
+## パイプライン出力
+
+| 出力ファイル | 説明 | 連携先スキル |
+|---|---|---|
+| `results/nhanes_data.csv` | NHANES 疫学データ | → epidemiology-public-health, survival-clinical |
+| `results/drug_mapping.json` | RxNorm 薬剤マッピング | → pharmacovigilance, pharmacogenomics |
+| `results/health_guidelines.json` | 臨床ガイドライン | → clinical-decision-support |
+| `results/health_disparities.csv` | 健康格差指標 | → epidemiology-public-health, causal-inference |
+
+## パイプライン統合
+
+```
+epidemiology-public-health ──→ public-health-data ──→ clinical-decision-support
+  (RR/OR/DAG)                  (NHANES/CDC/ODPHP)    (GRADE エビデンス)
+                                      │
+                                      ├──→ pharmacovigilance (RxNorm + 安全性)
+                                      ├──→ pharmacogenomics (RxNorm + PGx)
+                                      └──→ survival-clinical (NHANES コホート)
+```

package/src/.github/skills/scientific-regulatory-genomics/SKILL.md

@@ -0,0 +1,274 @@
+---
+name: scientific-regulatory-genomics
+description: |
+  レギュラトリーゲノミクススキル。RegulomeDB バリアント制御機能スコア、
+  ReMap 転写因子結合マッピング、4D Nucleome (4DN) 三次元ゲノム構造
+  解析の統合パイプライン。
+---
+
+# Scientific Regulatory Genomics
+
+RegulomeDB / ReMap / 4D Nucleome を統合した
+レギュラトリーゲノミクス (制御領域バリアント解析) パイプラインを提供する。
+
+## When to Use
+
+- 非コード領域バリアントの制御機能を評価するとき
+- RegulomeDB で SNP の調節的影響をスコアリングするとき
+- ReMap で転写因子結合部位のマッピングを確認するとき
+- 4DN データから三次元ゲノム構造 (TAD/ループ) を解析するとき
+- GWAS ヒットの制御メカニズムを解明するとき
+
+---
+
+## Quick Start
+
+## 1. RegulomeDB バリアント制御スコア
+
+```python
+import requests
+import pandas as pd
+
+REGULOMEDB_API = "https://regulomedb.org/regulome-search"
+
+
+def score_regulome_variants(variants):
+    """
+    RegulomeDB — 非コード領域バリアントの制御機能スコアリング。
+
+    Parameters:
+        variants: list — バリアントリスト (rsID or chr:pos 形式)
+            e.g., ["rs12345", "chr1:109274570"]
+
+    ToolUniverse:
+        RegulomeDB_score_variant(variant=variant)
+    """
+    results = []
+    for variant in variants:
+        params = {"regions": variant, "genome": "GRCh38", "format": "json"}
+        resp = requests.get(REGULOMEDB_API, params=params)
+        if resp.status_code != 200:
+            results.append({"variant": variant, "score": None, "error": True})
+            continue
+
+        data = resp.json()
+        for hit in data.get("@graph", []):
+            results.append({
+                "variant": variant,
+                "regulome_score": hit.get("regulome_score", {}).get("ranking", ""),
+                "probability": hit.get("regulome_score", {}).get("probability", ""),
+                "chrom": hit.get("chrom", ""),
+                "start": hit.get("start", ""),
+                "end": hit.get("end", ""),
+                "dnase": hit.get("dnase", ""),
+                "proteins_binding": hit.get("proteins_binding", []),
+                "motifs": hit.get("motifs", []),
+                "eqtls": hit.get("eqtls", []),
+                "chromatin_state": hit.get("chromatin_state", {}),
+            })
+
+    df = pd.DataFrame(results)
+    if not df.empty and "regulome_score" in df.columns:
+        high_func = (df["regulome_score"].astype(str).str.match(r"^[12]")).sum()
+        print(f"RegulomeDB: {len(variants)} variants scored, "
+              f"{high_func} with high regulatory function (score 1-2)")
+    return df
+```
+
+## 2. ReMap 転写因子結合マッピング
+
+```python
+REMAP_API = "https://remap.univ-amu.fr/api/v1"
+
+
+def search_remap_binding(chrom, start, end, genome="hg38"):
+    """
+    ReMap — ゲノム領域の転写因子/コレギュレーター結合マッピング。
+
+    Parameters:
+        chrom: str — 染色体 (e.g., "chr1")
+        start: int — 開始座標
+        end: int — 終了座標
+        genome: str — ゲノムアセンブリ ("hg38", "hg19", "mm10")
+
+    ToolUniverse:
+        ReMap_search_peaks(chrom=chrom, start=start, end=end)
+        ReMap_get_tf_targets(tf_name=tf_name)
+    """
+    params = {
+        "chrom": chrom,
+        "start": start,
+        "end": end,
+        "genome": genome,
+    }
+    resp = requests.get(f"{REMAP_API}/peaks/search", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for peak in data.get("peaks", []):
+        results.append({
+            "tf_name": peak.get("tf_name", ""),
+            "biotype": peak.get("biotype", ""),
+            "cell_type": peak.get("cell_type", ""),
+            "experiment": peak.get("experiment_accession", ""),
+            "peak_start": peak.get("start", ""),
+            "peak_end": peak.get("end", ""),
+            "score": peak.get("score", ""),
+        })
+
+    df = pd.DataFrame(results)
+    unique_tfs = df["tf_name"].nunique() if not df.empty else 0
+    print(f"ReMap {chrom}:{start}-{end}: {len(df)} peaks, {unique_tfs} unique TFs")
+    return df
+
+
+def get_remap_tf_targets(tf_name, genome="hg38"):
+    """
+    ReMap — 特定転写因子の全結合部位取得。
+
+    Parameters:
+        tf_name: str — 転写因子名 (e.g., "TP53", "CTCF", "STAT3")
+    """
+    params = {"tf": tf_name, "genome": genome}
+    resp = requests.get(f"{REMAP_API}/peaks/tf", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for peak in data.get("peaks", [])[:1000]:  # Limit for large TFs
+        results.append({
+            "chrom": peak.get("chrom", ""),
+            "start": peak.get("start", ""),
+            "end": peak.get("end", ""),
+            "cell_type": peak.get("cell_type", ""),
+            "score": peak.get("score", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"ReMap TF '{tf_name}': {len(df)} binding sites")
+    return df
+```
+
+## 3. 4D Nucleome (4DN) 三次元ゲノム構造
+
+```python
+FOURDN_API = "https://data.4dnucleome.org"
+
+
+def search_4dn_experiments(query, experiment_type=None):
+    """
+    4D Nucleome ポータル — 三次元ゲノム実験データ検索。
+
+    Parameters:
+        query: str — 検索クエリ (細胞株名、タンパク質名等)
+        experiment_type: str — 実験タイプ ("in situ Hi-C", "SPRITE", "GAM")
+
+    ToolUniverse:
+        FourDN_search_experiments(query=query)
+    """
+    params = {
+        "searchTerm": query,
+        "type": "ExperimentSetReplicate",
+        "format": "json",
+    }
+    if experiment_type:
+        params["experiment_type.display_title"] = experiment_type
+
+    resp = requests.get(f"{FOURDN_API}/search/", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("@graph", []):
+        results.append({
+            "accession": item.get("accession", ""),
+            "title": item.get("display_title", ""),
+            "experiment_type": item.get("experiment_type", {}).get("display_title", ""),
+            "biosource": item.get("biosource_summary", ""),
+            "lab": item.get("lab", {}).get("display_title", ""),
+            "status": item.get("status", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"4DN search '{query}': {len(df)} experiment sets")
+    return df
+```
+
+## 4. 制御バリアント統合解析パイプライン
+
+```python
+def regulatory_variant_pipeline(variants, genome="hg38"):
+    """
+    制御領域バリアント統合解析。
+
+    Parameters:
+        variants: list — バリアントリスト (rsID or chr:pos)
+    """
+    print("=" * 60)
+    print("Regulatory Variant Analysis Pipeline")
+    print("=" * 60)
+
+    # Step 1: RegulomeDB scoring
+    print("\n[1/3] RegulomeDB scoring...")
+    regulome_df = score_regulome_variants(variants)
+
+    # Step 2: ReMap TF binding for high-scoring variants
+    print("\n[2/3] ReMap TF binding analysis...")
+    remap_results = {}
+    for _, row in regulome_df.iterrows():
+        if row.get("chrom") and row.get("start"):
+            chrom = row["chrom"]
+            start = int(row["start"]) - 500
+            end = int(row["end"]) + 500
+            try:
+                remap_df = search_remap_binding(chrom, start, end, genome)
+                remap_results[row["variant"]] = remap_df
+            except Exception as e:
+                print(f"  ReMap error for {row['variant']}: {e}")
+
+    # Step 3: Summary
+    print("\n[3/3] Summary")
+    summary = {
+        "total_variants": len(variants),
+        "regulome_scored": len(regulome_df),
+        "high_regulatory": (
+            regulome_df["regulome_score"].astype(str).str.match(r"^[12]")
+        ).sum() if "regulome_score" in regulome_df.columns else 0,
+        "remap_annotated": len(remap_results),
+    }
+    print(f"  Total: {summary['total_variants']}, "
+          f"High regulatory: {summary['high_regulatory']}, "
+          f"ReMap annotated: {summary['remap_annotated']}")
+
+    return {"regulome": regulome_df, "remap": remap_results, "summary": summary}
+```
+
+---
+
+## 利用可能ツール
+
+| ToolUniverse カテゴリ | 主なツール |
+|---|---|
+| `regulomedb` | `RegulomeDB_score_variant` |
+| `remap` | `ReMap_search_peaks`, `ReMap_get_tf_targets` |
+| `fourdn_portal` | `FourDN_search_experiments` |
+
+## パイプライン出力
+
+| 出力ファイル | 説明 | 連携先スキル |
+|---|---|---|
+| `results/regulome_scores.csv` | バリアント制御スコア | → variant-interpretation, variant-effect-prediction |
+| `results/remap_binding.csv` | TF 結合マッピング | → epigenomics-chromatin, disease-research |
+| `results/4dn_contacts.json` | 3D ゲノム構造データ | → single-cell-genomics, epigenomics-chromatin |
+
+## パイプライン統合
+
+```
+variant-interpretation ──→ regulatory-genomics ──→ epigenomics-chromatin
+  (ACMG/AMP)               (RegulomeDB/ReMap/4DN)  (ChIP-seq/ATAC)
+                                    │
+                                    ├──→ disease-research (GWAS enhancer)
+                                    ├──→ gene-expression (eQTL/制御)
+                                    └──→ noncoding-rna (ncRNA 制御)
+```