@nahisaho/satori 0.16.0 → 0.18.0

package/src/.github/skills/scientific-pharmgkb-pgx/SKILL.md

@@ -0,0 +1,306 @@
+---
+name: scientific-pharmgkb-pgx
+description: |
+  PharmGKB 薬理ゲノミクススキル。PharmGKB REST API による
+  臨床アノテーション・薬物遺伝子関連・投与量ガイドライン・
+  スターアレル解析。ToolUniverse 連携: pharmgkb。
+tu_tools:
+  - key: pharmgkb
+    name: PharmGKB
+    description: 臨床アノテーション・薬物遺伝子関連・PGx ガイドライン
+---
+
+# Scientific PharmGKB PGx
+
+PharmGKB (Pharmacogenomics Knowledgebase) REST API を活用した
+薬理ゲノミクス臨床アノテーション・薬物遺伝子相互作用・投与量
+ガイドライン検索パイプラインを提供する。
+
+## When to Use
+
+- 薬物と遺伝子変異の関連を調べるとき
+- 臨床アノテーション (エビデンスレベル付き) を検索するとき
+- 投与量調整ガイドライン (CPIC/DPWG) を取得するとき
+- スターアレルと表現型の対応を確認するとき
+- 特定薬物の薬理ゲノミクス情報を包括的に取得するとき
+- 精密医療の薬物選択を支援するとき
+
+---
+
+## Quick Start
+
+## 1. 薬物・遺伝子検索
+
+```python
+import requests
+import pandas as pd
+
+PGKB_BASE = "https://api.pharmgkb.org/v1/data"
+
+
+def pharmgkb_search_drugs(query, limit=50):
+    """
+    PharmGKB — 薬物検索。
+
+    Parameters:
+        query: str — 薬物名 (例: "warfarin", "clopidogrel")
+        limit: int — 最大結果数
+    """
+    url = f"{PGKB_BASE}/chemical"
+    params = {"name": query, "view": "max"}
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("data", []):
+        results.append({
+            "pharmgkb_id": item.get("id", ""),
+            "name": item.get("name", ""),
+            "generic_names": "; ".join(
+                item.get("genericNames", [])),
+            "trade_names": "; ".join(
+                item.get("tradeNames", [])[:5]),
+            "type": item.get("type", ""),
+            "cross_references": len(
+                item.get("crossReferences", [])),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PharmGKB drugs: {len(df)} results "
+          f"(query='{query}')")
+    return df
+
+
+def pharmgkb_search_genes(query, limit=50):
+    """
+    PharmGKB — 遺伝子検索。
+
+    Parameters:
+        query: str — 遺伝子シンボル (例: "CYP2D6")
+        limit: int — 最大結果数
+    """
+    url = f"{PGKB_BASE}/gene"
+    params = {"symbol": query, "view": "max"}
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("data", []):
+        results.append({
+            "pharmgkb_id": item.get("id", ""),
+            "symbol": item.get("symbol", ""),
+            "name": item.get("name", ""),
+            "chromosome": item.get("chromosomeFormatted", ""),
+            "cpic_gene": item.get("cpicGene", False),
+            "has_prescribing_info": item.get(
+                "hasPrescribingInfo", False),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PharmGKB genes: {len(df)} results "
+          f"(query='{query}')")
+    return df
+```
+
+## 2. 臨床アノテーション取得
+
+```python
+def pharmgkb_clinical_annotations(gene_or_drug,
+                                     search_type="gene"):
+    """
+    PharmGKB — 臨床アノテーション検索。
+
+    Parameters:
+        gene_or_drug: str — 遺伝子シンボル or 薬物名
+        search_type: str — "gene" or "drug"
+    """
+    url = f"{PGKB_BASE}/clinicalAnnotation"
+    params = {"view": "max"}
+
+    if search_type == "gene":
+        # 遺伝子で検索
+        gene_url = f"{PGKB_BASE}/gene"
+        g_resp = requests.get(gene_url,
+                              params={"symbol": gene_or_drug},
+                              timeout=30)
+        g_resp.raise_for_status()
+        genes = g_resp.json().get("data", [])
+        if genes:
+            params["relatedGenes.id"] = genes[0].get("id", "")
+    else:
+        # 薬物で検索
+        drug_url = f"{PGKB_BASE}/chemical"
+        d_resp = requests.get(drug_url,
+                              params={"name": gene_or_drug},
+                              timeout=30)
+        d_resp.raise_for_status()
+        drugs = d_resp.json().get("data", [])
+        if drugs:
+            params["relatedChemicals.id"] = drugs[0].get("id", "")
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("data", []):
+        genes = [g.get("symbol", "")
+                 for g in item.get("relatedGenes", [])]
+        drugs = [c.get("name", "")
+                 for c in item.get("relatedChemicals", [])]
+        results.append({
+            "annotation_id": item.get("id", ""),
+            "level": item.get("level", ""),
+            "score": item.get("score", ""),
+            "genes": "; ".join(genes),
+            "drugs": "; ".join(drugs),
+            "phenotype_category": item.get(
+                "phenotypeCategory", ""),
+            "sentences": (item.get("textHtml") or "")[:300],
+        })
+
+    df = pd.DataFrame(results)
+    if not df.empty:
+        df = df.sort_values("level")
+
+    print(f"PharmGKB annotations: {len(df)} "
+          f"({search_type}={gene_or_drug})")
+    return df
+```
+
+## 3. 投与量ガイドライン取得
+
+```python
+def pharmgkb_dosing_guidelines(drug_name=None, gene=None):
+    """
+    PharmGKB — 投与量ガイドライン (CPIC/DPWG) 検索。
+
+    Parameters:
+        drug_name: str — 薬物名
+        gene: str — 遺伝子シンボル
+    """
+    url = f"{PGKB_BASE}/guideline"
+    params = {"view": "max"}
+
+    if drug_name:
+        params["relatedChemicals.name"] = drug_name
+    if gene:
+        params["relatedGenes.symbol"] = gene
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("data", []):
+        genes = [g.get("symbol", "")
+                 for g in item.get("relatedGenes", [])]
+        drugs = [c.get("name", "")
+                 for c in item.get("relatedChemicals", [])]
+        results.append({
+            "guideline_id": item.get("id", ""),
+            "name": item.get("name", ""),
+            "source": item.get("source", ""),
+            "genes": "; ".join(genes),
+            "drugs": "; ".join(drugs),
+            "recommendation": (item.get("textHtml") or "")[:500],
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PharmGKB guidelines: {len(df)} "
+          f"(drug={drug_name}, gene={gene})")
+    return df
+```
+
+## 4. PharmGKB 統合パイプライン
+
+```python
+def pharmgkb_pipeline(drug_name, genes=None,
+                         output_dir="results"):
+    """
+    PharmGKB 統合パイプライン。
+
+    Parameters:
+        drug_name: str — 薬物名
+        genes: list[str] — 関連遺伝子リスト
+        output_dir: str — 出力ディレクトリ
+    """
+    from pathlib import Path
+    output_dir = Path(output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    # 1) 薬物検索
+    drugs = pharmgkb_search_drugs(drug_name)
+    drugs.to_csv(output_dir / "drugs.csv", index=False)
+
+    # 2) 薬物の臨床アノテーション
+    annotations = pharmgkb_clinical_annotations(
+        drug_name, search_type="drug")
+    annotations.to_csv(output_dir / "annotations.csv",
+                       index=False)
+
+    # 3) 投与量ガイドライン
+    guidelines = pharmgkb_dosing_guidelines(
+        drug_name=drug_name)
+    guidelines.to_csv(output_dir / "guidelines.csv",
+                      index=False)
+
+    # 4) 関連遺伝子解析
+    if genes:
+        gene_results = []
+        for g in genes:
+            try:
+                g_ann = pharmgkb_clinical_annotations(
+                    g, search_type="gene")
+                g_ann["query_gene"] = g
+                gene_results.append(g_ann)
+            except Exception:
+                continue
+        if gene_results:
+            gene_df = pd.concat(gene_results,
+                                ignore_index=True)
+            gene_df.to_csv(
+                output_dir / "gene_annotations.csv",
+                index=False)
+
+    print(f"PharmGKB pipeline: {output_dir}")
+    return {
+        "drugs": drugs,
+        "annotations": annotations,
+        "guidelines": guidelines,
+    }
+```
+
+---
+
+## ToolUniverse 連携
+
+| TU Key | ツール名 | 連携内容 |
+|--------|---------|---------|
+| `pharmgkb` | PharmGKB | 臨床アノテーション・薬物遺伝子・PGx ガイドライン |
+
+## パイプライン統合
+
+```
+pharmacogenomics → pharmgkb-pgx → clinical-decision-support
+  (PGx 解析全般)  (PharmGKB API)  (臨床意思決定)
+       │                │                ↓
+  drug-discovery ──────┘         precision-oncology
+  (薬物開発)         │            (精密腫瘍学)
+                     ↓
+           variant-interpretation
+           (変異臨床解釈)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/drugs.csv` | 薬物情報 | → drug-discovery |
+| `results/annotations.csv` | 臨床アノテーション | → variant-interpretation |
+| `results/guidelines.csv` | 投与量ガイドライン | → clinical-decision-support |
+| `results/gene_annotations.csv` | 遺伝子別アノテーション | → pharmacogenomics |

package/src/.github/skills/scientific-semantic-scholar/SKILL.md

@@ -0,0 +1,298 @@
+---
+name: scientific-semantic-scholar
+description: |
+  Semantic Scholar 学術グラフスキル。Semantic Scholar Academic
+  Graph API による論文検索・著者プロファイル・引用グラフ・
+  推薦・TLDR 要約。ToolUniverse 連携: semantic_scholar。
+tu_tools:
+  - key: semantic_scholar
+    name: Semantic Scholar
+    description: 学術論文検索・引用解析・著者プロファイル
+---
+
+# Scientific Semantic Scholar
+
+Semantic Scholar Academic Graph API を活用した学術論文検索・
+引用ネットワーク解析・著者プロファイル・論文推薦パイプライン
+を提供する。
+
+## When to Use
+
+- 学術論文を高精度で検索するとき
+- 引用・被引用ネットワークを解析するとき
+- 著者の h-index・論文数・研究領域を調べるとき
+- 関連論文の推薦を受けるとき
+- TLDR (自動要約) を取得するとき
+- 特定分野の引用傾向を分析するとき
+- PubMed/OpenAlex 以外の学術検索エンジンを使うとき
+
+---
+
+## Quick Start
+
+## 1. 論文検索
+
+```python
+import requests
+import pandas as pd
+
+S2_BASE = "https://api.semanticscholar.org/graph/v1"
+S2_HEADERS = {}  # API key: {"x-api-key": "YOUR_KEY"}
+
+
+def semantic_scholar_search(query, limit=50,
+                              year_range=None,
+                              fields_of_study=None):
+    """
+    Semantic Scholar — 論文検索。
+
+    Parameters:
+        query: str — 検索クエリ
+        limit: int — 最大結果数
+        year_range: str — 年範囲 (例: "2020-2024")
+        fields_of_study: list[str] — 分野フィルタ
+    """
+    url = f"{S2_BASE}/paper/search"
+    params = {
+        "query": query,
+        "limit": min(limit, 100),
+        "fields": ("paperId,title,year,citationCount,"
+                    "influentialCitationCount,authors,"
+                    "journal,tldr,openAccessPdf,fieldsOfStudy"),
+    }
+    if year_range:
+        params["year"] = year_range
+    if fields_of_study:
+        params["fieldsOfStudy"] = ",".join(fields_of_study)
+
+    resp = requests.get(url, params=params,
+                        headers=S2_HEADERS, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for p in data.get("data", []):
+        authors = [a.get("name", "") for a in p.get("authors", [])]
+        tldr_text = ""
+        if p.get("tldr"):
+            tldr_text = p["tldr"].get("text", "")
+        results.append({
+            "paper_id": p.get("paperId", ""),
+            "title": p.get("title", ""),
+            "year": p.get("year"),
+            "citation_count": p.get("citationCount", 0),
+            "influential_citations": p.get(
+                "influentialCitationCount", 0),
+            "authors": "; ".join(authors[:5]),
+            "journal": (p.get("journal") or {}).get("name", ""),
+            "fields": ", ".join(p.get("fieldsOfStudy") or []),
+            "tldr": tldr_text[:300],
+            "pdf_url": (p.get("openAccessPdf") or {}).get("url", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"Semantic Scholar: {len(df)} papers "
+          f"(query='{query}')")
+    return df
+
+
+def semantic_scholar_get_paper(paper_id):
+    """
+    Semantic Scholar — 論文詳細取得。
+
+    Parameters:
+        paper_id: str — S2 Paper ID / DOI / ArXiv ID
+    """
+    url = f"{S2_BASE}/paper/{paper_id}"
+    params = {
+        "fields": ("paperId,title,year,abstract,citationCount,"
+                    "influentialCitationCount,authors,references,"
+                    "citations,journal,tldr,openAccessPdf,"
+                    "fieldsOfStudy,publicationDate,venue"),
+    }
+    resp = requests.get(url, params=params,
+                        headers=S2_HEADERS, timeout=30)
+    resp.raise_for_status()
+    return resp.json()
+```
+
+## 2. 著者プロファイル・引用解析
+
+```python
+def semantic_scholar_author(author_id, paper_limit=100):
+    """
+    Semantic Scholar — 著者プロファイル取得。
+
+    Parameters:
+        author_id: str — S2 Author ID
+        paper_limit: int — 取得論文数上限
+    """
+    url = f"{S2_BASE}/author/{author_id}"
+    params = {
+        "fields": ("authorId,name,affiliations,homepage,"
+                    "paperCount,citationCount,hIndex"),
+    }
+    resp = requests.get(url, params=params,
+                        headers=S2_HEADERS, timeout=30)
+    resp.raise_for_status()
+    profile = resp.json()
+
+    # 論文一覧
+    papers_url = f"{S2_BASE}/author/{author_id}/papers"
+    p_params = {
+        "fields": "paperId,title,year,citationCount,venue",
+        "limit": min(paper_limit, 1000),
+    }
+    p_resp = requests.get(papers_url, params=p_params,
+                          headers=S2_HEADERS, timeout=30)
+    p_resp.raise_for_status()
+
+    papers = []
+    for p in p_resp.json().get("data", []):
+        papers.append({
+            "paper_id": p.get("paperId", ""),
+            "title": p.get("title", ""),
+            "year": p.get("year"),
+            "citations": p.get("citationCount", 0),
+            "venue": p.get("venue", ""),
+        })
+
+    papers_df = pd.DataFrame(papers)
+
+    print(f"Author {profile.get('name', '')}: "
+          f"h-index={profile.get('hIndex', 0)}, "
+          f"{profile.get('paperCount', 0)} papers, "
+          f"{profile.get('citationCount', 0)} citations")
+    return profile, papers_df
+```
+
+## 3. 引用ネットワーク・影響度分析
+
+```python
+def semantic_scholar_citation_graph(paper_id,
+                                       direction="both",
+                                       limit=100):
+    """
+    Semantic Scholar — 引用グラフ取得。
+
+    Parameters:
+        paper_id: str — S2 Paper ID
+        direction: str — "citations", "references", "both"
+        limit: int — 各方向の上限
+    """
+    graphs = {}
+    fields = "paperId,title,year,citationCount,authors"
+
+    if direction in ("citations", "both"):
+        url = f"{S2_BASE}/paper/{paper_id}/citations"
+        resp = requests.get(url, params={"fields": fields,
+                            "limit": limit},
+                            headers=S2_HEADERS, timeout=30)
+        resp.raise_for_status()
+        cites = []
+        for c in resp.json().get("data", []):
+            cp = c.get("citingPaper", {})
+            cites.append({
+                "paper_id": cp.get("paperId", ""),
+                "title": cp.get("title", ""),
+                "year": cp.get("year"),
+                "citations": cp.get("citationCount", 0),
+            })
+        graphs["citations"] = pd.DataFrame(cites)
+
+    if direction in ("references", "both"):
+        url = f"{S2_BASE}/paper/{paper_id}/references"
+        resp = requests.get(url, params={"fields": fields,
+                            "limit": limit},
+                            headers=S2_HEADERS, timeout=30)
+        resp.raise_for_status()
+        refs = []
+        for r in resp.json().get("data", []):
+            rp = r.get("citedPaper", {})
+            refs.append({
+                "paper_id": rp.get("paperId", ""),
+                "title": rp.get("title", ""),
+                "year": rp.get("year"),
+                "citations": rp.get("citationCount", 0),
+            })
+        graphs["references"] = pd.DataFrame(refs)
+
+    for k, v in graphs.items():
+        print(f"  {k}: {len(v)} papers")
+    return graphs
+```
+
+## 4. 学術文献統合パイプライン
+
+```python
+def semantic_scholar_pipeline(query, year_range=None,
+                                 output_dir="results"):
+    """
+    Semantic Scholar 統合パイプライン。
+
+    Parameters:
+        query: str — 検索クエリ
+        year_range: str — 年範囲
+        output_dir: str — 出力ディレクトリ
+    """
+    from pathlib import Path
+    output_dir = Path(output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    # 1) 論文検索
+    papers = semantic_scholar_search(query,
+                                     year_range=year_range)
+    papers.to_csv(output_dir / "papers.csv", index=False)
+
+    # 2) トップ被引用論文の引用グラフ
+    if not papers.empty:
+        top = papers.sort_values("citation_count",
+                                  ascending=False).iloc[0]
+        pid = top["paper_id"]
+        graphs = semantic_scholar_citation_graph(pid)
+        for k, df in graphs.items():
+            df.to_csv(output_dir / f"{k}.csv", index=False)
+
+    # 3) 年次引用傾向
+    if not papers.empty and "year" in papers.columns:
+        yearly = papers.groupby("year").agg(
+            papers_count=("paper_id", "count"),
+            total_citations=("citation_count", "sum"),
+            avg_citations=("citation_count", "mean"),
+        ).reset_index()
+        yearly.to_csv(output_dir / "yearly_trend.csv",
+                      index=False)
+
+    print(f"Semantic Scholar pipeline: {output_dir}")
+    return {"papers": papers}
+```
+
+---
+
+## ToolUniverse 連携
+
+| TU Key | ツール名 | 連携内容 |
+|--------|---------|---------|
+| `semantic_scholar` | Semantic Scholar | 論文検索・引用解析・著者・TLDR |
+
+## パイプライン統合
+
+```
+literature-search → semantic-scholar → deep-research
+  (PubMed/NCBI)   (Academic Graph API) (knowledge synthesis)
+       │                  │                    ↓
+  crossref-metadata ─────┘             citation-checker
+  (DOI/metadata)     │                 (引用品質検証)
+                     ↓
+            gene-expression-transcriptomics
+            (論文引用データからの解析)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/papers.csv` | 論文検索結果 | → deep-research |
+| `results/citations.csv` | 被引用論文 | → citation-checker |
+| `results/references.csv` | 引用論文 | → meta-analysis |
+| `results/yearly_trend.csv` | 年次引用傾向 | → bibliometrics |