@nahisaho/satori 0.16.0 → 0.18.0

package/src/.github/skills/scientific-gwas-catalog/SKILL.md

@@ -0,0 +1,267 @@
+---
+name: scientific-gwas-catalog
+description: |
+  GWAS カタログスキル。NHGRI-EBI GWAS Catalog REST API によるゲノム
+  ワイド関連研究メタデータ・関連シグナル・形質・遺伝子座検索。
+  ToolUniverse 連携: gwas。
+tu_tools:
+  - key: gwas
+    name: GWAS Catalog
+    description: GWAS 関連シグナル・形質・遺伝子座検索
+---
+
+# Scientific GWAS Catalog
+
+NHGRI-EBI GWAS Catalog REST API を活用した GWAS メタデータ
+解析・遺伝子座レベル解釈パイプラインを提供する。
+
+## When to Use
+
+- GWAS Catalog から疾患/形質の関連バリアントを検索するとき
+- 遺伝的関連シグナルのエフェクトサイズ・P値を取得するとき
+- 特定遺伝子座の LD ブロック情報を解析するとき
+- 多形質 PheWAS-like 解析を実施するとき
+- GWAS サマリ統計量を下流解析に準備するとき
+- 公開 GWAS データから PRS ウェイトを抽出するとき
+
+---
+
+## Quick Start
+
+## 1. GWAS 関連シグナル検索
+
+```python
+import requests
+import pandas as pd
+import numpy as np
+
+GWAS_BASE = "https://www.ebi.ac.uk/gwas/rest/api"
+
+
+def gwas_search_associations(trait=None, gene=None, variant=None,
+                               p_upper=5e-8, limit=100):
+    """
+    GWAS Catalog — 関連シグナル検索。
+
+    Parameters:
+        trait: str — 形質/疾患 EFO ID or 名前 (例: "EFO_0001645")
+        gene: str — 遺伝子名 (例: "BRCA1")
+        variant: str — rsID (例: "rs1234567")
+        p_upper: float — P値上限
+        limit: int — 最大結果数
+    """
+    if trait:
+        url = f"{GWAS_BASE}/efoTraits/{trait}/associations"
+    elif gene:
+        url = f"{GWAS_BASE}/associations/search/findByGene"
+    elif variant:
+        url = f"{GWAS_BASE}/singleNucleotidePolymorphisms/{variant}/associations"
+    else:
+        url = f"{GWAS_BASE}/associations"
+
+    params = {"size": limit}
+    if gene:
+        params["geneName"] = gene
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    associations = data.get("_embedded", {}).get("associations", [])
+    results = []
+    for assoc in associations:
+        p_value = assoc.get("pvalue", 1.0)
+        if p_value and float(p_value) > p_upper:
+            continue
+
+        loci = assoc.get("loci", [{}])
+        genes = []
+        for locus in loci:
+            for gene_info in locus.get("authorReportedGenes", []):
+                genes.append(gene_info.get("geneName", ""))
+
+        snps = []
+        for snp_info in assoc.get("snps", []):
+            snps.append(snp_info.get("rsId", ""))
+
+        results.append({
+            "association_id": assoc.get("associationId", ""),
+            "p_value": float(p_value) if p_value else None,
+            "p_value_mlog": assoc.get("pvalueMantissa", 0),
+            "or_beta": assoc.get("orPerCopyNum", None),
+            "beta_num": assoc.get("betaNum", None),
+            "beta_direction": assoc.get("betaDirection", ""),
+            "ci": assoc.get("range", ""),
+            "risk_allele_freq": assoc.get("riskFrequency", ""),
+            "snps": "; ".join(snps),
+            "genes": "; ".join(genes),
+            "trait": assoc.get("efoTraits", [{}])[0].get("trait", "")
+                     if assoc.get("efoTraits") else "",
+            "study_accession": assoc.get("study", {}).get(
+                "accessionId", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"GWAS associations: {len(df)} results "
+          f"(trait={trait}, gene={gene}, p<{p_upper})")
+    return df.sort_values("p_value") if not df.empty else df
+```
+
+## 2. GWAS 研究メタデータ検索
+
+```python
+def gwas_search_studies(query=None, efo_trait=None, limit=50):
+    """
+    GWAS Catalog — 研究メタデータ検索。
+
+    Parameters:
+        query: str — フリーテキスト検索
+        efo_trait: str — EFO 形質 ID
+        limit: int — 最大結果数
+    """
+    if efo_trait:
+        url = f"{GWAS_BASE}/efoTraits/{efo_trait}/studies"
+    else:
+        url = f"{GWAS_BASE}/studies/search/findByDiseaseTrait"
+
+    params = {"size": limit}
+    if query:
+        params["diseaseTrait"] = query
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    studies = data.get("_embedded", {}).get("studies", [])
+    results = []
+    for s in studies:
+        results.append({
+            "accession": s.get("accessionId", ""),
+            "title": s.get("title", ""),
+            "pubmed_id": s.get("publicationInfo", {}).get(
+                "pubmedId", ""),
+            "author": s.get("publicationInfo", {}).get(
+                "author", {}).get("fullname", ""),
+            "journal": s.get("publicationInfo", {}).get(
+                "publication", ""),
+            "date": s.get("publicationInfo", {}).get(
+                "publicationDate", ""),
+            "initial_sample_size": s.get("initialSampleSize", ""),
+            "replication_sample_size": s.get(
+                "replicationSampleSize", ""),
+            "ancestry": s.get("ancestries", []),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"GWAS studies: {len(df)} results")
+    return df
+```
+
+## 3. GWAS 形質検索・PheWAS
+
+```python
+def gwas_phewas(variant_rsid, p_threshold=5e-8):
+    """
+    GWAS Catalog — バリアント PheWAS (形質横断検索)。
+
+    Parameters:
+        variant_rsid: str — rsID (例: "rs7903146")
+        p_threshold: float — P値閾値
+    """
+    url = (f"{GWAS_BASE}/singleNucleotidePolymorphisms/"
+           f"{variant_rsid}/associations")
+    resp = requests.get(url, params={"size": 500}, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    associations = data.get("_embedded", {}).get("associations", [])
+    results = []
+    for assoc in associations:
+        p_val = assoc.get("pvalue", 1.0)
+        if p_val and float(p_val) > p_threshold:
+            continue
+        for trait in assoc.get("efoTraits", []):
+            results.append({
+                "variant": variant_rsid,
+                "trait": trait.get("trait", ""),
+                "efo_uri": trait.get("shortForm", ""),
+                "p_value": float(p_val) if p_val else None,
+                "or_beta": assoc.get("orPerCopyNum", None),
+                "study": assoc.get("study", {}).get(
+                    "accessionId", ""),
+            })
+
+    df = pd.DataFrame(results)
+    if not df.empty:
+        df = df.sort_values("p_value")
+    print(f"PheWAS {variant_rsid}: {len(df)} trait associations")
+    return df
+```
+
+## 4. GWAS 統合パイプライン
+
+```python
+def gwas_catalog_pipeline(trait_query, output_dir="results"):
+    """
+    GWAS Catalog 統合パイプライン。
+
+    Parameters:
+        trait_query: str — 形質/疾患名
+        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) 研究検索
+    studies = gwas_search_studies(query=trait_query)
+    studies.to_csv(output_dir / "gwas_studies.csv", index=False)
+
+    # 2) 関連シグナル
+    assocs = gwas_search_associations(gene=None, trait=None)
+    assocs.to_csv(output_dir / "gwas_associations.csv", index=False)
+
+    # 3) トップバリアントの PheWAS
+    if not assocs.empty:
+        top_snps = assocs["snps"].str.split("; ").explode().unique()[:5]
+        phewas_all = []
+        for rsid in top_snps:
+            if rsid.startswith("rs"):
+                phewas = gwas_phewas(rsid)
+                phewas_all.append(phewas)
+        if phewas_all:
+            phewas_df = pd.concat(phewas_all, ignore_index=True)
+            phewas_df.to_csv(output_dir / "phewas.csv", index=False)
+
+    print(f"GWAS pipeline: {output_dir}")
+    return {"studies": studies, "associations": assocs}
+```
+
+---
+
+## ToolUniverse 連携
+
+| TU Key | ツール名 | 連携内容 |
+|--------|---------|---------|
+| `gwas` | GWAS Catalog | 関連シグナル・形質・研究メタデータ検索 |
+
+## パイプライン統合
+
+```
+disease-research → gwas-catalog → variant-interpretation
+  (DisGeNET/OMIM)  (GWAS Catalog)  (ACMG/AMP)
+       │                 │                ↓
+  population-genetics ──┘         variant-effect-prediction
+  (Fst/PCA)           │          (CADD/SpliceAI)
+                       ↓
+                  precision-oncology
+                  (臨床的意義判定)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/gwas_studies.csv` | GWAS 研究メタデータ | → literature-search |
+| `results/gwas_associations.csv` | 関連シグナル | → variant-interpretation |
+| `results/phewas.csv` | PheWAS 結果 | → disease-research |

package/src/.github/skills/scientific-human-cell-atlas/SKILL.md

@@ -0,0 +1,294 @@
+---
+name: scientific-human-cell-atlas
+description: |
+  Human Cell Atlas (HCA) データポータルスキル。HCA Data Portal API
+  プロジェクト検索・ファイルダウンロード・CELLxGENE Census 統合・
+  細胞型アノテーション・アトラス構築。ToolUniverse 連携: hca_tools。
+tu_tools:
+  - key: hca_tools
+    name: Human Cell Atlas Tools
+    description: HCA データポータル プロジェクト・バンドル・ファイル検索
+---
+
+# Scientific Human Cell Atlas
+
+HCA Data Portal / CELLxGENE Census を活用した大規模シングルセル
+アトラスデータアクセス・解析パイプラインを提供する。
+
+## When to Use
+
+- HCA Data Portal からプロジェクト・実験データを検索するとき
+- CELLxGENE Census で大規模シングルセルアトラスを照会するとき
+- 特定組織/疾患の細胞型構成を調べるとき
+- 複数 HCA プロジェクト間で細胞型を比較するとき
+- シングルセルアトラスのリファレンスマッピングを行うとき
+- 希少細胞型の発見・アノテーションを実施するとき
+
+---
+
+## Quick Start
+
+## 1. HCA Data Portal プロジェクト検索
+
+```python
+import requests
+import pandas as pd
+import json
+
+HCA_BASE = "https://service.azul.data.humancellatlas.org"
+HCA_CATALOG = "dcp44"
+
+
+def hca_search_projects(keyword=None, organ=None, disease=None,
+                         species="Homo sapiens", limit=25):
+    """
+    HCA Data Portal — プロジェクト検索。
+
+    Parameters:
+        keyword: str — キーワード検索
+        organ: str — 臓器 (例: "lung", "heart")
+        disease: str — 疾患 (例: "COVID-19")
+        species: str — 生物種
+        limit: int — 最大結果数
+    """
+    url = f"{HCA_BASE}/index/projects"
+    params = {"catalog": HCA_CATALOG, "size": limit}
+
+    filters = {}
+    if organ:
+        filters["organ"] = {"is": [organ]}
+    if disease:
+        filters["disease"] = {"is": [disease]}
+    if species:
+        filters["genusSpecies"] = {"is": [species]}
+    if keyword:
+        params["q"] = keyword
+    if filters:
+        params["filters"] = json.dumps(filters)
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    projects = []
+    for hit in data.get("hits", []):
+        proj = hit.get("projects", [{}])[0]
+        samples = hit.get("samples", [{}])[0]
+        protocols = hit.get("protocols", [{}])[0]
+        projects.append({
+            "project_id": proj.get("projectId", ""),
+            "title": proj.get("projectTitle", ""),
+            "organ": ", ".join(samples.get("organ", [])),
+            "disease": ", ".join(samples.get("disease", [])),
+            "species": ", ".join(samples.get("genusSpecies", [])),
+            "cell_count": hit.get("cellSuspensions", [{}])[0].get(
+                "totalCells", 0),
+            "library_method": ", ".join(protocols.get(
+                "libraryConstructionApproach", [])),
+            "donor_count": samples.get("donorCount", 0),
+        })
+
+    df = pd.DataFrame(projects)
+    print(f"HCA: {len(df)} projects found")
+    return df
+```
+
+## 2. HCA ファイル取得
+
+```python
+def hca_get_project_files(project_id, file_format=None):
+    """
+    HCA — プロジェクトのファイル一覧取得。
+
+    Parameters:
+        project_id: str — プロジェクト UUID
+        file_format: str — ファイル形式 (例: "h5ad", "loom", "csv")
+    """
+    url = f"{HCA_BASE}/index/files"
+    filters = {"projectId": {"is": [project_id]}}
+    if file_format:
+        filters["fileFormat"] = {"is": [file_format]}
+
+    params = {
+        "catalog": HCA_CATALOG,
+        "filters": json.dumps(filters),
+        "size": 100,
+    }
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    files = []
+    for hit in data.get("hits", []):
+        for f in hit.get("files", []):
+            files.append({
+                "file_id": f.get("uuid", ""),
+                "name": f.get("name", ""),
+                "format": f.get("format", ""),
+                "size_bytes": f.get("size", 0),
+                "url": f.get("url", ""),
+            })
+
+    df = pd.DataFrame(files)
+    print(f"HCA files ({project_id[:8]}): {len(df)} files")
+    return df
+```
+
+## 3. CELLxGENE Census アトラスクエリ
+
+```python
+def cellxgene_census_query(organism="homo_sapiens", tissue=None,
+                            disease=None, cell_type=None,
+                            gene_list=None, max_cells=50000):
+    """
+    CELLxGENE Census — 大規模シングルセルアトラスクエリ。
+
+    Parameters:
+        organism: str — 生物種
+        tissue: str — 組織
+        disease: str — 疾患
+        cell_type: str — 細胞型
+        gene_list: list[str] — 取得遺伝子リスト
+        max_cells: int — 最大細胞数
+    """
+    import cellxgene_census
+
+    census = cellxgene_census.open_soma()
+
+    # 観察フィルタ構築
+    obs_filters = []
+    if tissue:
+        obs_filters.append(f"tissue == '{tissue}'")
+    if disease:
+        obs_filters.append(f"disease == '{disease}'")
+    if cell_type:
+        obs_filters.append(f"cell_type == '{cell_type}'")
+
+    obs_filter = " and ".join(obs_filters) if obs_filters else None
+
+    # 遺伝子フィルタ
+    var_filter = None
+    if gene_list:
+        genes_str = "', '".join(gene_list)
+        var_filter = f"feature_name in ['{genes_str}']"
+
+    adata = cellxgene_census.get_anndata(
+        census,
+        organism=organism,
+        obs_value_filter=obs_filter,
+        var_value_filter=var_filter,
+        obs_column_names=[
+            "cell_type", "tissue", "disease",
+            "donor_id", "dataset_id", "assay",
+        ],
+    )
+
+    if adata.n_obs > max_cells:
+        import numpy as np
+        idx = np.random.choice(adata.n_obs, max_cells, replace=False)
+        adata = adata[idx].copy()
+
+    census.close()
+    print(f"CELLxGENE Census: {adata.n_obs} cells × {adata.n_vars} genes")
+    return adata
+```
+
+## 4. 細胞型構成解析
+
+```python
+import scanpy as sc
+
+
+def cell_type_composition(adata, groupby="tissue", cell_type_col="cell_type"):
+    """
+    細胞型構成の定量比較。
+
+    Parameters:
+        adata: AnnData — シングルセルデータ
+        groupby: str — グループ変数
+        cell_type_col: str — 細胞型カラム名
+    """
+    # 構成比計算
+    composition = (
+        adata.obs.groupby([groupby, cell_type_col])
+        .size()
+        .unstack(fill_value=0)
+    )
+    composition_pct = composition.div(composition.sum(axis=1), axis=0) * 100
+
+    print(f"Cell type composition: {composition.shape[0]} groups × "
+          f"{composition.shape[1]} cell types")
+    return composition_pct
+```
+
+## 5. HCA 統合パイプライン
+
+```python
+def hca_atlas_pipeline(organ, disease=None, output_dir="results"):
+    """
+    HCA + CELLxGENE 統合アトラスパイプライン。
+
+    Parameters:
+        organ: str — 対象臓器
+        disease: str — 対象疾患
+        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) HCA プロジェクト検索
+    projects = hca_search_projects(organ=organ, disease=disease)
+    projects.to_csv(output_dir / "hca_projects.csv", index=False)
+
+    # 2) CELLxGENE Census クエリ
+    adata = cellxgene_census_query(tissue=organ, disease=disease)
+
+    # 3) 前処理
+    sc.pp.normalize_total(adata, target_sum=1e4)
+    sc.pp.log1p(adata)
+    sc.pp.highly_variable_genes(adata, n_top_genes=2000)
+    adata = adata[:, adata.var["highly_variable"]].copy()
+    sc.pp.pca(adata)
+    sc.pp.neighbors(adata)
+    sc.tl.umap(adata)
+
+    # 4) 細胞型構成
+    composition = cell_type_composition(adata)
+    composition.to_csv(output_dir / "cell_type_composition.csv")
+
+    # 5) 保存
+    adata.write(output_dir / "hca_atlas.h5ad")
+
+    print(f"HCA atlas pipeline: {output_dir}")
+    return {"projects": projects, "adata": adata, "composition": composition}
+```
+
+---
+
+## ToolUniverse 連携
+
+| TU Key | ツール名 | 連携内容 |
+|--------|---------|---------|
+| `hca_tools` | HCA Tools | プロジェクト検索・ファイルダウンロード |
+
+## パイプライン統合
+
+```
+single-cell-genomics → human-cell-atlas → scvi-integration
+  (scanpy 標準)         (HCA/CELLxGENE)   (scVI 統合)
+       │                      │                ↓
+  spatial-transcriptomics ───┘         cell-type-annotation
+  (Visium/MERFISH)       │              (リファレンスマッピング)
+                         ↓
+                   gpu-singlecell
+                   (大規模処理)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/hca_projects.csv` | HCA プロジェクト一覧 | → single-cell-genomics |
+| `results/hca_atlas.h5ad` | アトラス AnnData | → scvi-integration |
+| `results/cell_type_composition.csv` | 細胞型構成比 | → spatial-transcriptomics |