@nahisaho/satori 0.12.0 → 0.14.0

package/src/.github/skills/scientific-ontology-enrichment/SKILL.md

@@ -0,0 +1,340 @@
+---
+name: scientific-ontology-enrichment
+description: |
+  オントロジー・エンリッチメント解析スキル。EFO 実験ファクターオントロジー、
+  OLS オントロジー検索サービス、Enrichr 遺伝子セット濃縮解析、
+  UMLS メタシソーラス統一医学言語体系の統合パイプライン。
+---
+
+# Scientific Ontology Enrichment
+
+EFO / OLS / Enrichr / UMLS を統合した
+オントロジー検索・エンリッチメント解析パイプラインを提供する。
+
+## When to Use
+
+- EFO で実験条件 (疾患・細胞型・組織) のオントロジー ID を取得するとき
+- OLS で複数オントロジー横断検索 (HP, MONDO, DOID, GO, CHEBI) するとき
+- Enrichr で遺伝子リストの濃縮解析を行うとき
+- UMLS CUI で異なる用語体系間のマッピングを行うとき
+- GWAS Catalog の trait を EFO 用語で標準化するとき
+
+---
+
+## Quick Start
+
+## 1. EFO 実験ファクターオントロジー
+
+```python
+import requests
+import pandas as pd
+
+OLS_API = "https://www.ebi.ac.uk/ols4/api"
+
+
+def search_efo(query, exact=False):
+    """
+    EFO (Experimental Factor Ontology) 検索。
+
+    Parameters:
+        query: str — 検索語 (疾患名、細胞型、組織名等)
+        exact: bool — 完全一致検索
+
+    ToolUniverse:
+        EFO_search(query=query, exact=exact)
+    """
+    params = {
+        "q": query,
+        "ontology": "efo",
+        "exact": str(exact).lower(),
+        "rows": 30,
+    }
+    resp = requests.get(f"{OLS_API}/search", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for doc in data.get("response", {}).get("docs", []):
+        results.append({
+            "efo_id": doc.get("obo_id", ""),
+            "label": doc.get("label", ""),
+            "description": (doc.get("description") or [""])[0][:200],
+            "iri": doc.get("iri", ""),
+            "ontology": doc.get("ontology_name", ""),
+            "is_defining_ontology": doc.get("is_defining_ontology", False),
+            "synonyms": doc.get("synonym", []),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"EFO search '{query}': {len(df)} terms")
+    return df
+```
+
+## 2. OLS マルチオントロジー検索
+
+```python
+def search_ols(query, ontologies=None, type_filter=None):
+    """
+    OLS (Ontology Lookup Service) マルチオントロジー横断検索。
+
+    Parameters:
+        query: str — 検索語
+        ontologies: list — オントロジー ID リスト (e.g., ["hp", "mondo", "go"])
+        type_filter: str — "class", "property", "individual"
+
+    ToolUniverse:
+        OLS_search(query=query, ontology=ontology)
+        OLS_get_term(ontology=ontology, iri=iri)
+        OLS_get_ancestors(ontology=ontology, iri=iri)
+    """
+    params = {"q": query, "rows": 50}
+    if ontologies:
+        params["ontology"] = ",".join(ontologies)
+    if type_filter:
+        params["type"] = type_filter
+
+    resp = requests.get(f"{OLS_API}/search", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for doc in data.get("response", {}).get("docs", []):
+        results.append({
+            "obo_id": doc.get("obo_id", ""),
+            "label": doc.get("label", ""),
+            "ontology": doc.get("ontology_name", ""),
+            "description": (doc.get("description") or [""])[0][:200],
+            "iri": doc.get("iri", ""),
+            "synonyms": doc.get("synonym", []),
+            "has_children": doc.get("has_children", False),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"OLS search '{query}' "
+          f"[{','.join(ontologies) if ontologies else 'all'}]: "
+          f"{len(df)} terms")
+    return df
+
+
+def get_ols_term_hierarchy(ontology, term_id):
+    """
+    OLS 用語の階層構造 (ancestors/descendants) 取得。
+
+    Parameters:
+        ontology: str — オントロジー ID (e.g., "hp", "go")
+        term_id: str — OBO ID (e.g., "HP:0001250")
+    """
+    iri = f"http://purl.obolibrary.org/obo/{term_id.replace(':', '_')}"
+    encoded_iri = requests.utils.quote(requests.utils.quote(iri, safe=""), safe="")
+
+    # Ancestors
+    anc_resp = requests.get(
+        f"{OLS_API}/ontologies/{ontology}/terms/{encoded_iri}/ancestors"
+    )
+
+    # Descendants
+    desc_resp = requests.get(
+        f"{OLS_API}/ontologies/{ontology}/terms/{encoded_iri}/descendants"
+    )
+
+    hierarchy = {"ancestors": [], "descendants": []}
+
+    if anc_resp.status_code == 200:
+        for t in anc_resp.json().get("_embedded", {}).get("terms", []):
+            hierarchy["ancestors"].append({
+                "id": t.get("obo_id", ""),
+                "label": t.get("label", ""),
+            })
+
+    if desc_resp.status_code == 200:
+        for t in desc_resp.json().get("_embedded", {}).get("terms", []):
+            hierarchy["descendants"].append({
+                "id": t.get("obo_id", ""),
+                "label": t.get("label", ""),
+            })
+
+    print(f"OLS hierarchy {term_id}: "
+          f"{len(hierarchy['ancestors'])} ancestors, "
+          f"{len(hierarchy['descendants'])} descendants")
+    return hierarchy
+```
+
+## 3. Enrichr 遺伝子セット濃縮解析
+
+```python
+ENRICHR_API = "https://maayanlab.cloud/Enrichr"
+
+
+def run_enrichr(gene_list, description="", gene_set_libraries=None):
+    """
+    Enrichr 遺伝子リスト濃縮解析。
+
+    Parameters:
+        gene_list: list — 遺伝子シンボルリスト (e.g., ["TP53", "BRCA1", "EGFR"])
+        description: str — 解析の説明
+        gene_set_libraries: list — 使用する遺伝子セットライブラリ
+
+    ToolUniverse:
+        Enrichr_submit_gene_list(genes=gene_list)
+        Enrichr_get_enrichment(user_list_id=id, library=library)
+    """
+    if gene_set_libraries is None:
+        gene_set_libraries = [
+            "GO_Biological_Process_2023",
+            "GO_Molecular_Function_2023",
+            "KEGG_2021_Human",
+            "Reactome_2022",
+            "WikiPathway_2023_Human",
+            "DisGeNET",
+        ]
+
+    # Submit gene list
+    genes_str = "\n".join(gene_list)
+    submit_resp = requests.post(
+        f"{ENRICHR_API}/addList",
+        files={"list": (None, genes_str), "description": (None, description)},
+    )
+    submit_resp.raise_for_status()
+    user_list_id = submit_resp.json().get("userListId")
+    print(f"Enrichr: submitted {len(gene_list)} genes (ID={user_list_id})")
+
+    # Get enrichment results per library
+    all_results = {}
+    for library in gene_set_libraries:
+        enrich_resp = requests.get(
+            f"{ENRICHR_API}/enrich",
+            params={"userListId": user_list_id, "backgroundType": library},
+        )
+        enrich_resp.raise_for_status()
+        data = enrich_resp.json()
+
+        results = []
+        for term_data in data.get(library, []):
+            results.append({
+                "rank": term_data[0],
+                "term": term_data[1],
+                "p_value": term_data[2],
+                "z_score": term_data[3],
+                "combined_score": term_data[4],
+                "overlap_genes": term_data[5],
+                "adjusted_p": term_data[6],
+            })
+
+        df = pd.DataFrame(results)
+        if not df.empty:
+            df = df.sort_values("adjusted_p")
+        all_results[library] = df
+        sig_count = (df["adjusted_p"] < 0.05).sum() if not df.empty else 0
+        print(f"  {library}: {sig_count} significant terms (FDR < 0.05)")
+
+    return all_results
+```
+
+## 4. UMLS メタシソーラスマッピング
+
+```python
+UMLS_API = "https://uts-ws.nlm.nih.gov/rest"
+
+
+def search_umls(query, api_key, search_type="words"):
+    """
+    UMLS メタシソーラス検索。
+
+    Parameters:
+        query: str — 検索語 (疾患名、症状、薬剤名)
+        api_key: str — UMLS API キー
+        search_type: str — "words", "exact", "leftTruncation"
+
+    ToolUniverse:
+        UMLS_search(query=query, search_type=search_type)
+        UMLS_get_concept(cui=cui)
+    """
+    params = {
+        "string": query,
+        "searchType": search_type,
+        "apiKey": api_key,
+        "pageSize": 25,
+    }
+    resp = requests.get(f"{UMLS_API}/search/current", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data.get("result", {}).get("results", []):
+        results.append({
+            "cui": item.get("ui", ""),
+            "name": item.get("name", ""),
+            "root_source": item.get("rootSource", ""),
+            "uri": item.get("uri", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"UMLS search '{query}': {len(df)} concepts")
+    return df
+
+
+def get_umls_crosswalk(cui, api_key, target_source=None):
+    """
+    UMLS CUI からの用語体系間マッピング。
+
+    Parameters:
+        cui: str — UMLS CUI (e.g., "C0023264")
+        api_key: str — UMLS API キー
+        target_source: str — ターゲット用語体系 (e.g., "SNOMEDCT_US", "ICD10CM", "MeSH")
+    """
+    params = {"apiKey": api_key, "pageSize": 100}
+    if target_source:
+        params["sabs"] = target_source
+
+    resp = requests.get(f"{UMLS_API}/content/current/CUI/{cui}/atoms", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    mappings = []
+    for atom in data.get("result", []):
+        mappings.append({
+            "source": atom.get("rootSource", ""),
+            "code": atom.get("sourceConcept", ""),
+            "name": atom.get("name", ""),
+            "term_type": atom.get("termType", ""),
+        })
+
+    df = pd.DataFrame(mappings)
+    if target_source:
+        df = df[df["source"] == target_source]
+
+    print(f"UMLS crosswalk {cui}: {len(df)} mappings "
+          f"({target_source or 'all sources'})")
+    return df
+```
+
+---
+
+## 利用可能ツール
+
+| ToolUniverse カテゴリ | 主なツール |
+|---|---|
+| `efo` | `EFO_search` |
+| `ols` | `OLS_search`, `OLS_get_term`, `OLS_get_ancestors` |
+| `enrichr` | `Enrichr_submit_gene_list`, `Enrichr_get_enrichment` |
+| `umls` | `UMLS_search`, `UMLS_get_concept` |
+
+## パイプライン出力
+
+| 出力ファイル | 説明 | 連携先スキル |
+|---|---|---|
+| `results/efo_terms.csv` | EFO 標準化用語 | → disease-research, gene-expression |
+| `results/enrichr_results/` | 遺伝子セット濃縮結果 | → pathway-enrichment, multi-omics |
+| `results/umls_mapping.json` | UMLS 用語マッピング | → clinical-decision-support, public-health-data |
+| `results/ontology_hierarchy.json` | オントロジー階層 | → text-mining-nlp, knowledge-graph |
+
+## パイプライン統合
+
+```
+disease-research ──→ ontology-enrichment ──→ pathway-enrichment
+  (GWAS/DisGeNET)    (EFO/OLS/UMLS/Enrichr)   (KEGG/Reactome/GO)
+                              │
+                              ├──→ biothings-idmapping (CUI→Gene→Protein)
+                              ├──→ public-health-data (UMLS→RxNorm)
+                              └──→ clinical-reporting (SNOMED/ICD マッピング)
+```

package/src/.github/skills/scientific-pharmacology-targets/SKILL.md

@@ -0,0 +1,323 @@
+---
+name: scientific-pharmacology-targets
+description: |
+  薬理学的ターゲットプロファイリングスキル。BindingDB 結合親和性、
+  GPCRdb GPCR 構造-活性、GtoPdb 薬理学、BRENDA 酵素動態、
+  Pharos 未解明ターゲット(TDL)の統合解析パイプライン。
+---
+
+# Scientific Pharmacology Targets
+
+複数の薬理学データベース (BindingDB, GPCRdb, GtoPdb, BRENDA, Pharos) を
+統合した包括的ターゲットプロファイリングパイプラインを提供する。
+
+## When to Use
+
+- 特定タンパク質の既知リガンド・結合親和性を調べるとき
+- GPCR のリガンド・変異・構造情報を取得するとき
+- 薬物-ターゲット相互作用のデータベース横断検索を行うとき
+- 酵素阻害剤データ (BRENDA) を調べるとき
+- 未解明ターゲット (Tdark/Tbio) のドラッガビリティを評価するとき
+
+---
+
+## Quick Start
+
+## 1. BindingDB 結合親和性データ取得
+
+```python
+import requests
+import pandas as pd
+
+
+def get_bindingdb_ligands(uniprot_id, cutoff=None):
+    """
+    BindingDB から UniProt ID ベースのリガンド結合データを取得。
+
+    Parameters:
+        uniprot_id: str — UniProt accession (e.g., "P00533")
+        cutoff: float | None — affinity cutoff nM
+
+    ToolUniverse:
+        BindingDB_get_ligands_by_uniprot(uniprot_id=uniprot_id)
+        BindingDB_get_targets_by_compound(smiles=smiles)
+    """
+    url = "https://bindingdb.org/axis2/services/BDBService"
+    params = {
+        "uniprot": uniprot_id,
+        "response": "json",
+    }
+    if cutoff:
+        params["cutoff"] = cutoff
+
+    resp = requests.get(f"{url}/getLigandsByUniprot", params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    ligands = data.get("getLigandsByUniprotResponse", {}).get("affinities", [])
+    results = []
+    for lig in ligands:
+        results.append({
+            "monomer_id": lig.get("monomerid", ""),
+            "smiles": lig.get("smiles", ""),
+            "affinity_type": lig.get("affinity_type", ""),
+            "affinity_value_nm": lig.get("affinity", ""),
+            "source": lig.get("source", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"BindingDB '{uniprot_id}': {len(df)} ligands")
+    return df
+```
+
+## 2. GPCRdb GPCR プロファイリング
+
+```python
+def get_gpcrdb_profile(protein_entry):
+    """
+    GPCRdb から GPCR のリガンド・変異・構造情報を取得。
+
+    Parameters:
+        protein_entry: str — GPCRdb entry name (e.g., "adrb2_human")
+
+    ToolUniverse:
+        GPCRdb_get_protein(entry_name=protein_entry)
+        GPCRdb_get_ligands(entry_name=protein_entry)
+        GPCRdb_get_mutations(entry_name=protein_entry)
+        GPCRdb_get_structures(entry_name=protein_entry)
+        GPCRdb_list_proteins()
+    """
+    base = "https://gpcrdb.org/services"
+
+    # Protein info
+    resp_p = requests.get(f"{base}/protein/{protein_entry}/")
+    resp_p.raise_for_status()
+    protein = resp_p.json()
+
+    # Ligands
+    resp_l = requests.get(f"{base}/ligands/{protein_entry}/")
+    ligands = resp_l.json() if resp_l.status_code == 200 else []
+
+    # Mutations
+    resp_m = requests.get(f"{base}/mutants/{protein_entry}/")
+    mutations = resp_m.json() if resp_m.status_code == 200 else []
+
+    # Structures
+    resp_s = requests.get(f"{base}/structure/protein/{protein_entry}/")
+    structures = resp_s.json() if resp_s.status_code == 200 else []
+
+    profile = {
+        "entry_name": protein.get("entry_name", ""),
+        "name": protein.get("name", ""),
+        "family": protein.get("family", ""),
+        "species": protein.get("species", ""),
+        "num_ligands": len(ligands),
+        "num_mutations": len(mutations),
+        "num_structures": len(structures),
+    }
+
+    print(f"GPCRdb '{protein_entry}': "
+          f"{profile['num_ligands']} ligands, "
+          f"{profile['num_mutations']} mutations, "
+          f"{profile['num_structures']} structures")
+    return profile, ligands, mutations, structures
+```
+
+## 3. GtoPdb 薬理学データ
+
+```python
+def get_gtopdb_target_pharmacology(target_id):
+    """
+    Guide to PHARMACOLOGY (GtoPdb) から
+    ターゲットの薬理学的相互作用データを取得。
+
+    ToolUniverse:
+        GtoPdb_get_target(target_id=target_id)
+        GtoPdb_get_target_interactions(target_id=target_id)
+        GtoPdb_get_ligand(ligand_id=ligand_id)
+        GtoPdb_search_interactions(query=query)
+    """
+    base = "https://www.guidetopharmacology.org/services"
+
+    # Target info
+    resp_t = requests.get(f"{base}/targets/{target_id}")
+    resp_t.raise_for_status()
+    target = resp_t.json()
+
+    # Interactions
+    resp_i = requests.get(f"{base}/targets/{target_id}/interactions")
+    interactions = resp_i.json() if resp_i.status_code == 200 else []
+
+    results = []
+    for ix in interactions:
+        results.append({
+            "ligand_id": ix.get("ligandId", ""),
+            "ligand_name": ix.get("ligandName", ""),
+            "type": ix.get("type", ""),
+            "action": ix.get("action", ""),
+            "affinity_type": ix.get("affinityType", ""),
+            "affinity_median": ix.get("affinityMedian", ""),
+            "approved": ix.get("approvedDrug", False),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"GtoPdb target {target_id} ({target.get('name', '')}): "
+          f"{len(df)} interactions")
+    return target, df
+```
+
+## 4. Pharos/TCRD 未解明ターゲット検索
+
+```python
+def search_pharos_targets(query, tdl=None):
+    """
+    Pharos / TCRD からターゲット情報を取得。
+    Target Development Level (TDL) でフィルタ可能。
+
+    Parameters:
+        query: str — gene symbol or target name
+        tdl: str | None — "Tclin", "Tchem", "Tbio", "Tdark"
+
+    ToolUniverse:
+        Pharos_search_targets(q=query)
+        Pharos_get_target(q=query)
+        Pharos_get_tdl_summary()
+        Pharos_get_disease_targets(disease_name=disease_name)
+    """
+    url = "https://pharos-api.ncats.io/graphql"
+    gql = """
+    query TargetSearch($term: String!, $top: Int) {
+        targets(filter: { term: $term }, top: $top) {
+            targets {
+                name
+                sym
+                uniprot { accession }
+                tdl
+                fam
+                novelty
+                jensenScore
+                diseaseCount
+                ligandCount
+            }
+            count
+        }
+    }
+    """
+    variables = {"term": query, "top": 20}
+    resp = requests.post(url, json={"query": gql, "variables": variables})
+    resp.raise_for_status()
+    data = resp.json()["data"]["targets"]
+
+    results = []
+    for t in data["targets"]:
+        results.append({
+            "symbol": t.get("sym", ""),
+            "name": t.get("name", ""),
+            "uniprot": t.get("uniprot", {}).get("accession", ""),
+            "tdl": t.get("tdl", ""),
+            "family": t.get("fam", ""),
+            "novelty": t.get("novelty", 0),
+            "disease_count": t.get("diseaseCount", 0),
+            "ligand_count": t.get("ligandCount", 0),
+        })
+
+    df = pd.DataFrame(results)
+    if tdl:
+        df = df[df["tdl"] == tdl]
+
+    print(f"Pharos '{query}': {len(df)} targets"
+          f"{f' (TDL={tdl})' if tdl else ''}")
+    return df
+```
+
+## 5. 統合ターゲットプロファイリング
+
+```python
+def integrated_target_profile(uniprot_id, gene_symbol):
+    """
+    複数データベースを統合したターゲットプロファイル。
+
+    ToolUniverse (横断):
+        BindingDB_get_ligands_by_uniprot(uniprot_id)
+        Pharos_get_target(q=gene_symbol)
+        GtoPdb_get_targets() → GtoPdb_get_target_interactions()
+    """
+    profile = {
+        "uniprot_id": uniprot_id,
+        "gene_symbol": gene_symbol,
+    }
+
+    # BindingDB ligands
+    try:
+        bdb_df = get_bindingdb_ligands(uniprot_id)
+        profile["bindingdb_ligand_count"] = len(bdb_df)
+    except Exception:
+        profile["bindingdb_ligand_count"] = 0
+
+    # Pharos TDL
+    try:
+        pharos_df = search_pharos_targets(gene_symbol)
+        if not pharos_df.empty:
+            row = pharos_df.iloc[0]
+            profile["tdl"] = row.get("tdl", "")
+            profile["novelty"] = row.get("novelty", 0)
+    except Exception:
+        profile["tdl"] = "Unknown"
+
+    print(f"Integrated profile {gene_symbol}: TDL={profile.get('tdl', '?')}, "
+          f"BindingDB={profile['bindingdb_ligand_count']} ligands")
+    return profile
+```
+
+## References
+
+### Output Files
+
+| ファイル | 形式 |
+|---|---|
+| `results/bindingdb_ligands.csv` | CSV |
+| `results/gpcrdb_profile.json` | JSON |
+| `results/gtopdb_interactions.csv` | CSV |
+| `results/pharos_targets.csv` | CSV |
+| `results/integrated_target_profile.json` | JSON |
+
+### 利用可能ツール
+
+| カテゴリ | 主要ツール | 用途 |
+|---|---|---|
+| BindingDB | `BindingDB_get_ligands_by_uniprot` | UniProt→リガンド |
+| BindingDB | `BindingDB_get_ligands_by_uniprots` | バッチ |
+| BindingDB | `BindingDB_get_ligands_by_pdb` | PDB→リガンド |
+| BindingDB | `BindingDB_get_targets_by_compound` | 化合物→ターゲット |
+| GPCRdb | `GPCRdb_get_protein` | GPCR 詳細 |
+| GPCRdb | `GPCRdb_get_ligands` | GPCR リガンド |
+| GPCRdb | `GPCRdb_get_mutations` | GPCR 変異 |
+| GPCRdb | `GPCRdb_get_structures` | GPCR 構造 |
+| GPCRdb | `GPCRdb_list_proteins` | GPCR 一覧 |
+| GtoPdb | `GtoPdb_get_target` | ターゲット情報 |
+| GtoPdb | `GtoPdb_get_target_interactions` | 相互作用 |
+| GtoPdb | `GtoPdb_get_ligand` | リガンド情報 |
+| GtoPdb | `GtoPdb_get_targets` | ターゲット一覧 |
+| GtoPdb | `GtoPdb_list_ligands` | リガンド一覧 |
+| GtoPdb | `GtoPdb_get_disease` | 疾患関連 |
+| GtoPdb | `GtoPdb_list_diseases` | 疾患一覧 |
+| GtoPdb | `GtoPdb_search_interactions` | 相互作用検索 |
+| BRENDA | `BRENDA_get_inhibitors` | 酵素阻害剤 |
+| Pharos | `Pharos_search_targets` | ターゲット検索 |
+| Pharos | `Pharos_get_target` | ターゲット詳細 |
+| Pharos | `Pharos_get_tdl_summary` | TDL サマリー |
+| Pharos | `Pharos_get_disease_targets` | 疾患→ターゲット |
+
+### 参照スキル
+
+| スキル | 関連 |
+|---|---|
+| `scientific-drug-target-interaction` | DTI 予測 |
+| `scientific-compound-similarity` | 化合物類似性 |
+| `scientific-compound-screening` | 化合物スクリーニング |
+| `scientific-molecular-docking` | 分子ドッキング |
+| `scientific-protein-interaction-network` | PPI ネットワーク |
+
+### 依存パッケージ
+
+`requests`, `pandas`