@nahisaho/satori 0.14.0 → 0.16.0

package/src/.github/skills/scientific-string-network-api/SKILL.md

@@ -0,0 +1,376 @@
+---
+name: scientific-string-network-api
+description: |
+  STRING/BioGRID/STITCH ネットワーク解析スキル。STRING タンパク質相互作用
+  ネットワーク直接 API、BioGRID 実験的 PPI、STITCH 化学-タンパク質ネットワーク、
+  ネットワークトポロジー解析・コミュニティ検出・機能濃縮統合パイプライン。
+---
+
+# Scientific STRING Network API
+
+STRING v12 / BioGRID / STITCH API を活用した PPI・化合物-タンパク質
+ネットワーク解析パイプラインを提供する。既存の protein-interaction-network
+スキル (IntAct/HumanBase) を補完し、STRING 直接 API ベースの高度な
+ネットワーク分析を統合。
+
+## When to Use
+
+- STRING API でタンパク質相互作用ネットワークを直接構築するとき
+- BioGRID から実験的エビデンスベースの PPI を取得するとき
+- STITCH で化合物-タンパク質間ネットワークを検索するとき
+- ネットワークトポロジー指標 (次数分布・媒介中心性) を計算するとき
+- PPI ネットワーク上でコミュニティ検出を行うとき
+- 機能濃縮解析 (STRING enrichment) をネットワーク上で実行するとき
+
+---
+
+## Quick Start
+
+## 1. STRING PPI ネットワーク取得
+
+```python
+import requests
+import pandas as pd
+import networkx as nx
+
+STRING_API = "https://string-db.org/api"
+OUTPUT_FORMAT = "json"
+
+
+def get_string_network(proteins, species=9606, score_threshold=400,
+                       network_type="functional", limit=50):
+    """
+    STRING PPI ネットワーク取得。
+
+    Parameters:
+        proteins: list — タンパク質名リスト (例: ["TP53", "MDM2", "BRCA1"])
+        species: int — NCBI Taxonomy ID (9606=human)
+        score_threshold: int — 信頼スコア閾値 (0-1000)
+        network_type: str — "functional" or "physical"
+        limit: int — interactor 最大数
+
+    ToolUniverse:
+        STRING_get_protein_interactions(
+            protein_ids=proteins, species=species,
+            confidence_score=score_threshold/1000,
+            network_type=network_type, limit=limit
+        )
+    """
+    url = f"{STRING_API}/{OUTPUT_FORMAT}/network"
+    params = {
+        "identifiers": "\r".join(proteins),
+        "species": species,
+        "required_score": score_threshold,
+        "network_type": network_type,
+        "limit": limit,
+    }
+
+    resp = requests.post(url, data=params)
+    resp.raise_for_status()
+    interactions = resp.json()
+
+    rows = []
+    for i in interactions:
+        rows.append({
+            "protein_a": i.get("preferredName_A"),
+            "protein_b": i.get("preferredName_B"),
+            "combined_score": i.get("score"),
+            "nscore": i.get("nscore"),
+            "fscore": i.get("fscore"),
+            "pscore": i.get("pscore"),
+            "ascore": i.get("ascore"),
+            "escore": i.get("escore"),
+            "dscore": i.get("dscore"),
+            "tscore": i.get("tscore"),
+        })
+
+    df = pd.DataFrame(rows)
+    print(f"STRING network: {len(df)} interactions "
+          f"(score ≥ {score_threshold/1000})")
+    return df
+```
+
+## 2. BioGRID 実験的 PPI 取得
+
+```python
+def get_biogrid_interactions(genes, organism=9606, evidence_type=None,
+                             api_key="YOUR_KEY", limit=500):
+    """
+    BioGRID 実験的 PPI データ取得。
+
+    Parameters:
+        genes: list — 遺伝子名リスト
+        organism: int — NCBI Taxonomy ID
+        evidence_type: str — "physical" or "genetic"
+        api_key: str — BioGRID API key (https://webservice.thebiogrid.org)
+        limit: int — 最大取得数
+
+    ToolUniverse:
+        BioGRID_get_interactions(
+            gene_names=genes, organism=organism,
+            interaction_type=evidence_type, limit=limit
+        )
+    """
+    url = "https://webservice.thebiogrid.org/interactions"
+    params = {
+        "accessKey": api_key,
+        "geneList": "|".join(genes),
+        "organism": organism,
+        "format": "json",
+        "max": limit,
+        "searchNames": "true",
+        "includeInteractors": "true",
+    }
+    if evidence_type:
+        params["interSpeciesExcluded"] = "true"
+
+    resp = requests.get(url, params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    rows = []
+    for _, interaction in data.items():
+        rows.append({
+            "gene_a": interaction.get("OFFICIAL_SYMBOL_A"),
+            "gene_b": interaction.get("OFFICIAL_SYMBOL_B"),
+            "experimental_system": interaction.get("EXPERIMENTAL_SYSTEM"),
+            "throughput": interaction.get("THROUGHPUT"),
+            "pubmed_id": interaction.get("PUBMED_ID"),
+            "source_db": "BioGRID",
+        })
+
+    df = pd.DataFrame(rows)
+    print(f"BioGRID: {len(df)} interactions for {genes}")
+    return df
+```
+
+## 3. STITCH 化合物-タンパク質ネットワーク
+
+```python
+def get_stitch_interactions(identifiers, species=9606, score=400, limit=20):
+    """
+    STITCH 化合物-タンパク質相互作用取得。
+
+    Parameters:
+        identifiers: list — CID (化合物) または遺伝子名リスト
+        species: int — NCBI Taxonomy ID
+        score: int — 信頼スコア閾値
+        limit: int — 最大結果数
+
+    ToolUniverse:
+        STITCH_get_chemical_protein_interactions(
+            identifiers=identifiers, species=species,
+            required_score=score, limit=limit
+        )
+        STITCH_get_interaction_partners(identifiers=identifiers)
+        STITCH_resolve_identifier(identifiers=identifiers)
+    """
+    url = f"https://stitch.embl.de/api/{OUTPUT_FORMAT}/interactionsList"
+    params = {
+        "identifiers": "\r".join(identifiers),
+        "species": species,
+        "required_score": score,
+        "limit": limit,
+    }
+
+    resp = requests.post(url, data=params)
+    resp.raise_for_status()
+    interactions = resp.json()
+
+    rows = []
+    for i in interactions:
+        rows.append({
+            "interactor_a": i.get("preferredName_A", i.get("stringId_A")),
+            "interactor_b": i.get("preferredName_B", i.get("stringId_B")),
+            "combined_score": i.get("score"),
+            "is_chemical": "CID" in str(i.get("stringId_A", ""))
+                or "CID" in str(i.get("stringId_B", "")),
+        })
+
+    df = pd.DataFrame(rows)
+    print(f"STITCH: {len(df)} chemical-protein interactions")
+    return df
+```
+
+## 4. ネットワーク構築 & トポロジー解析
+
+```python
+def build_network(interaction_df, source_col="protein_a", target_col="protein_b",
+                  weight_col="combined_score"):
+    """
+    NetworkX グラフ構築 & トポロジー解析。
+
+    Parameters:
+        interaction_df: DataFrame — 相互作用データ
+        source_col, target_col: str — ノードカラム名
+        weight_col: str — エッジ重みカラム名
+    """
+    G = nx.Graph()
+    for _, row in interaction_df.iterrows():
+        G.add_edge(
+            row[source_col], row[target_col],
+            weight=row.get(weight_col, 1.0),
+        )
+
+    # トポロジー指標
+    degree = dict(G.degree())
+    betweenness = nx.betweenness_centrality(G)
+    closeness = nx.closeness_centrality(G)
+    clustering = nx.clustering(G)
+
+    metrics = pd.DataFrame({
+        "node": list(degree.keys()),
+        "degree": list(degree.values()),
+        "betweenness": [betweenness[n] for n in degree],
+        "closeness": [closeness[n] for n in degree],
+        "clustering": [clustering[n] for n in degree],
+    }).sort_values("betweenness", ascending=False)
+
+    print(f"Network: {G.number_of_nodes()} nodes, "
+          f"{G.number_of_edges()} edges, "
+          f"density={nx.density(G):.4f}")
+    return G, metrics
+```
+
+## 5. コミュニティ検出
+
+```python
+from networkx.algorithms.community import greedy_modularity_communities
+
+
+def detect_communities(G, resolution=1.0):
+    """
+    ネットワーク上のコミュニティ (モジュール) 検出。
+
+    Parameters:
+        G: nx.Graph — ネットワークグラフ
+        resolution: float — 解像度パラメータ
+    """
+    communities = list(greedy_modularity_communities(G, resolution=resolution))
+    modularity = nx.algorithms.community.modularity(G, communities)
+
+    comm_data = []
+    for i, comm in enumerate(communities):
+        for node in comm:
+            comm_data.append({"node": node, "community": i})
+
+    df = pd.DataFrame(comm_data)
+    print(f"Communities: {len(communities)} detected, "
+          f"modularity={modularity:.4f}")
+    return df, modularity
+```
+
+## 6. STRING 機能濃縮解析
+
+```python
+def string_enrichment(proteins, species=9606):
+    """
+    STRING API 機能濃縮解析 (GO/KEGG/Reactome/InterPro)。
+
+    Parameters:
+        proteins: list — タンパク質名リスト
+        species: int — NCBI Taxonomy ID
+    """
+    url = f"{STRING_API}/{OUTPUT_FORMAT}/enrichment"
+    params = {
+        "identifiers": "\r".join(proteins),
+        "species": species,
+    }
+
+    resp = requests.post(url, data=params)
+    resp.raise_for_status()
+    enrichment = resp.json()
+
+    rows = []
+    for e in enrichment:
+        rows.append({
+            "category": e.get("category"),
+            "term": e.get("term"),
+            "description": e.get("description"),
+            "p_value": e.get("p_value"),
+            "fdr": e.get("fdr"),
+            "number_of_genes": e.get("number_of_genes"),
+            "input_genes": e.get("inputGenes", ""),
+        })
+
+    df = pd.DataFrame(rows)
+    if not df.empty:
+        df = df.sort_values("fdr")
+    print(f"Enrichment: {len(df)} terms, "
+          f"{df[df['fdr'] < 0.05].shape[0]} significant (FDR<0.05)")
+    return df
+```
+
+## 7. 統合 PPI 解析パイプライン
+
+```python
+def integrated_ppi_pipeline(genes, species=9606, score=700):
+    """
+    STRING + BioGRID + STITCH 統合 PPI パイプライン。
+
+    Pipeline:
+        STRING network → BioGRID validation → topology → communities →
+        enrichment
+    """
+    # STRING ネットワーク
+    string_df = get_string_network(genes, species, score)
+
+    # ネットワーク構築 & トポロジー
+    G, metrics = build_network(string_df)
+
+    # コミュニティ検出
+    comm_df, modularity = detect_communities(G)
+
+    # STRING 濃縮解析
+    all_nodes = list(G.nodes())
+    enrichment = string_enrichment(all_nodes[:500], species)
+
+    result = {
+        "n_nodes": G.number_of_nodes(),
+        "n_edges": G.number_of_edges(),
+        "density": round(nx.density(G), 4),
+        "n_communities": comm_df["community"].nunique(),
+        "modularity": round(modularity, 4),
+        "hub_genes": metrics.head(10)["node"].tolist(),
+        "n_enriched_terms": len(enrichment[enrichment["fdr"] < 0.05])
+            if not enrichment.empty else 0,
+    }
+
+    print(f"\n=== Integrated PPI Pipeline ===")
+    print(f"Nodes: {result['n_nodes']}, Edges: {result['n_edges']}")
+    print(f"Hub genes: {', '.join(result['hub_genes'][:5])}")
+    return result
+```
+
+---
+
+## パイプライン統合
+
+```
+drug-target-profiling → string-network-api → pathway-enrichment
+  (候補ターゲット)       (STRING PPI 構築)     (GO/KEGG 濃縮)
+        │                       │                    ↓
+protein-interaction ───┘        │           ontology-enrichment
+  (IntAct/HumanBase)            ↓             (EFO/Enrichr)
+                       network-analysis
+                       (既存スキル補完)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/string_network.csv` | STRING PPI ネットワーク | → network-analysis |
+| `results/ppi_topology.csv` | トポロジー指標 | → drug-target-profiling |
+| `results/ppi_communities.csv` | コミュニティ割当 | → pathway-enrichment |
+| `results/string_enrichment.csv` | 機能濃縮結果 | → ontology-enrichment |
+
+## 利用可能ツール (ToolUniverse SMCP)
+
+| ツール名 | 用途 |
+|---------|------|
+| `STRING_get_protein_interactions` | STRING PPI 取得 |
+| `BioGRID_get_interactions` | BioGRID 実験的 PPI |
+| `STITCH_get_chemical_protein_interactions` | STITCH 化合物-タンパク質 |
+| `STITCH_get_interaction_partners` | STITCH 相互作用パートナー |
+| `STITCH_resolve_identifier` | STITCH ID 解決 |

package/src/.github/skills/scientific-toxicology-env/SKILL.md

@@ -0,0 +1,309 @@
+---
+name: scientific-toxicology-env
+description: |
+  毒性学・環境衛生スキル。CTD (Comparative Toxicogenomics Database)
+  化学-遺伝子-疾患関連・ToxCast/Tox21 高スループット毒性スクリーニング・
+  IRIS ヒトリスク評価・T3DB 食品/環境毒性物質・PubChem BioAssay 毒性データ。
+---
+
+# Scientific Toxicology & Environmental Health
+
+CTD / Tox21 / ToxCast / T3DB / IRIS を活用した毒性学・環境衛生
+パイプラインを提供する。化学物質-遺伝子-疾患関連、高スループット
+毒性スクリーニング、ヒト健康リスク評価。
+
+## When to Use
+
+- 化学物質が関連する遺伝子・疾患を調べるとき (CTD)
+- Tox21/ToxCast アッセイデータで毒性メカニズムを解析するとき
+- IRIS リスク評価データ (RfD/RfC/UR) を参照するとき
+- 食品・環境毒性物質の詳細データベース (T3DB) を検索するとき
+- PubChem BioAssay から毒性関連ハイスループットデータを取得するとき
+- ADMET 毒性予測と実験毒性データを併用するとき
+
+---
+
+## Quick Start
+
+## 1. CTD 化学-遺伝子-疾患関連検索
+
+```python
+import requests
+import pandas as pd
+import json
+
+CTD_BASE = "https://ctdbase.org/tools"
+
+
+def ctd_chemical_gene(chemical_name, limit=100):
+    """
+    CTD — 化学物質-遺伝子相互作用を検索。
+
+    Parameters:
+        chemical_name: str — 化学物質名 (例: "Bisphenol A")
+        limit: int — 最大取得数
+    """
+    url = f"{CTD_BASE}/batchQuery.go"
+    params = {
+        "inputType": "chem",
+        "inputTerms": chemical_name,
+        "report": "genes_curated",
+        "format": "json",
+    }
+    resp = requests.get(url, params=params, timeout=60)
+    resp.raise_for_status()
+    data = resp.json()[:limit]
+
+    results = []
+    for entry in data:
+        results.append({
+            "chemical": entry.get("ChemicalName", ""),
+            "gene": entry.get("GeneSymbol", ""),
+            "organism": entry.get("Organism", ""),
+            "interaction": entry.get("Interaction", ""),
+            "pubmed_ids": entry.get("PubMedIDs", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"CTD: {chemical_name} → {len(df)} gene interactions")
+    return df
+```
+
+## 2. CTD 化学-疾患関連検索
+
+```python
+def ctd_chemical_disease(chemical_name, limit=100):
+    """
+    CTD — 化学物質-疾患関連を検索。
+
+    Parameters:
+        chemical_name: str — 化学物質名
+        limit: int — 最大取得数
+    """
+    url = f"{CTD_BASE}/batchQuery.go"
+    params = {
+        "inputType": "chem",
+        "inputTerms": chemical_name,
+        "report": "diseases_curated",
+        "format": "json",
+    }
+    resp = requests.get(url, params=params, timeout=60)
+    resp.raise_for_status()
+    data = resp.json()[:limit]
+
+    results = []
+    for entry in data:
+        results.append({
+            "chemical": entry.get("ChemicalName", ""),
+            "disease": entry.get("DiseaseName", ""),
+            "disease_id": entry.get("DiseaseID", ""),
+            "direct_evidence": entry.get("DirectEvidence", ""),
+            "inference_score": float(entry.get("InferenceScore", 0)),
+        })
+
+    df = pd.DataFrame(results)
+    df = df.sort_values("inference_score", ascending=False)
+    print(f"CTD: {chemical_name} → {len(df)} disease associations")
+    return df
+```
+
+## 3. Tox21/ToxCast アッセイデータ取得
+
+```python
+COMPTOX_BASE = "https://comptox.epa.gov/dashboard/api"
+
+
+def tox21_assay_search(chemical_identifier, assay_source="Tox21"):
+    """
+    CompTox Dashboard — Tox21/ToxCast アッセイ結果を取得。
+
+    Parameters:
+        chemical_identifier: str — DTXSID or CAS
+        assay_source: str — "Tox21" or "ToxCast"
+    """
+    # CompTox Dashboard API で化学物質のアッセイデータ取得
+    url = f"{COMPTOX_BASE}/chemical/search"
+    params = {"query": chemical_identifier}
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    chem_data = resp.json()
+
+    dtxsid = chem_data.get("dtxsid", chemical_identifier)
+
+    # アッセイエンドポイント取得
+    url_assay = f"{COMPTOX_BASE}/chemical/{dtxsid}/assays"
+    resp_assay = requests.get(url_assay, timeout=30)
+    resp_assay.raise_for_status()
+    assays = resp_assay.json()
+
+    results = []
+    for assay in assays:
+        if assay_source.lower() in assay.get("assaySource", "").lower():
+            results.append({
+                "assay_name": assay.get("assayName", ""),
+                "assay_source": assay.get("assaySource", ""),
+                "endpoint": assay.get("assayEndpoint", ""),
+                "activity": assay.get("activity", ""),
+                "ac50_um": assay.get("ac50", None),
+                "hit_call": assay.get("hitCall", ""),
+            })
+
+    df = pd.DataFrame(results)
+    n_active = (df["hit_call"] == "Active").sum() if len(df) > 0 else 0
+    print(f"Tox21/ToxCast: {dtxsid} → {len(df)} assays, {n_active} active")
+    return df
+```
+
+## 4. T3DB 毒性物質検索
+
+```python
+T3DB_BASE = "https://t3db.ca/api"
+
+
+def t3db_search(query, search_type="name"):
+    """
+    T3DB — 食品/環境毒性物質データベース検索。
+
+    Parameters:
+        query: str — 検索語
+        search_type: str — "name", "cas", "category"
+    """
+    url = f"{T3DB_BASE}/toxins/search"
+    params = {"query": query, "search_type": search_type}
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for toxin in data.get("toxins", []):
+        results.append({
+            "name": toxin.get("name", ""),
+            "t3db_id": toxin.get("t3db_id", ""),
+            "cas_number": toxin.get("cas_number", ""),
+            "category": toxin.get("category", ""),
+            "toxicity_class": toxin.get("toxicity_class", ""),
+            "ld50_oral": toxin.get("ld50_oral", ""),
+            "target_organs": toxin.get("target_organs", []),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"T3DB: '{query}' → {len(df)} toxins")
+    return df
+```
+
+## 5. EPA IRIS リスク評価
+
+```python
+def iris_risk_assessment(chemical_name):
+    """
+    EPA IRIS — ヒト健康リスク評価データ取得。
+
+    Parameters:
+        chemical_name: str — 化学物質名
+    """
+    url = "https://iris.epa.gov/AtoZ"
+    resp = requests.get(url, timeout=30)
+
+    # IRIS は構造化 API なし — スクレイピングまたはローカルデータ
+    # 代替: CompTox Dashboard API 経由
+    url_comptox = f"{COMPTOX_BASE}/chemical/search"
+    params = {"query": chemical_name}
+    resp = requests.get(url_comptox, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    risk_data = {
+        "chemical": chemical_name,
+        "dtxsid": data.get("dtxsid", ""),
+        "rfd_oral_mg_kg_day": data.get("rfdOral", None),
+        "rfc_inhalation_mg_m3": data.get("rfcInhalation", None),
+        "cancer_classification": data.get("cancerClassification", ""),
+        "oral_slope_factor": data.get("oralSlopeFactor", None),
+        "inhalation_unit_risk": data.get("inhalationUnitRisk", None),
+    }
+
+    print(f"IRIS: {chemical_name}")
+    for k, v in risk_data.items():
+        if v and k != "chemical":
+            print(f"  {k}: {v}")
+    return risk_data
+```
+
+## 6. 毒性パスウェイ解析
+
+```python
+def toxicity_pathway_analysis(chemical_name, species="Homo sapiens"):
+    """
+    CTD + パスウェイ統合毒性解析。
+
+    Parameters:
+        chemical_name: str — 化学物質名
+        species: str — 生物種
+    """
+    # 1) CTD 遺伝子取得
+    gene_df = ctd_chemical_gene(chemical_name, limit=500)
+    if species:
+        gene_df = gene_df[gene_df["organism"] == species]
+
+    gene_list = gene_df["gene"].unique().tolist()
+
+    # 2) CTD 疾患取得
+    disease_df = ctd_chemical_disease(chemical_name, limit=100)
+
+    # 3) パスウェイ濃縮 (KEGG enrichment via Enrichr)
+    enrichr_url = "https://maayanlab.cloud/Enrichr"
+    add_resp = requests.post(
+        f"{enrichr_url}/addList",
+        files={"list": (None, "\n".join(gene_list))}
+    )
+    user_list_id = add_resp.json()["userListId"]
+
+    enrich_resp = requests.get(
+        f"{enrichr_url}/enrich",
+        params={"userListId": user_list_id, "backgroundType": "KEGG_2021_Human"}
+    )
+    pathways = enrich_resp.json().get("KEGG_2021_Human", [])
+
+    pathway_results = []
+    for pw in pathways[:20]:
+        pathway_results.append({
+            "pathway": pw[1],
+            "p_value": pw[2],
+            "adj_p_value": pw[6],
+            "genes": pw[5],
+        })
+
+    print(f"Toxicity pathway: {chemical_name}")
+    print(f"  Target genes: {len(gene_list)}")
+    print(f"  Diseases: {len(disease_df)}")
+    print(f"  Pathways: {len(pathway_results)}")
+    return {
+        "genes": gene_df,
+        "diseases": disease_df,
+        "pathways": pd.DataFrame(pathway_results),
+    }
+```
+
+---
+
+## パイプライン統合
+
+```
+admet-pharmacokinetics → toxicology-env → pharmacovigilance
+  (ADMET 毒性予測)       (CTD/Tox21/IRIS)   (市販後安全性)
+        │                       │                  ↓
+cheminformatics ───────────────┘           disease-research
+  (RDKit 構造アラート)    │                (疾患-遺伝子関連)
+                          ↓
+                   public-health-data
+                   (CDC/WHO 公衆衛生)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/ctd_gene_interactions.csv` | CTD 化学-遺伝子関連 | → pathway-enrichment |
+| `results/ctd_disease_associations.csv` | CTD 化学-疾患関連 | → disease-research |
+| `results/tox21_assays.csv` | Tox21/ToxCast アッセイ結果 | → admet-pharmacokinetics |
+| `results/toxicity_pathways.json` | 毒性パスウェイ解析結果 | → pharmacovigilance |