@nahisaho/satori 0.15.0 → 0.17.0

package/src/.github/skills/scientific-nci60-screening/SKILL.md

@@ -0,0 +1,307 @@
+---
+name: scientific-nci60-screening
+description: |
+  NCI-60 がん細胞株薬剤応答スキル。CellMiner API 薬剤感受性・
+  NCI-60 GI50/LC50 データ・DepMap cancer dependency 統合・
+  薬剤-分子マーカー相関・細胞株パネル比較解析。
+---
+
+# Scientific NCI-60 Screening
+
+CellMiner / NCI-60 / DepMap を活用したがん細胞株薬剤応答
+パイプラインを提供する。高スループットスクリーニングデータの
+統合解析、薬剤感受性マーカー同定、パネル比較。
+
+## When to Use
+
+- NCI-60 細胞株パネルの薬剤応答 (GI50) を解析するとき
+- CellMiner から化合物活性データを取得するとき
+- 薬剤感受性と分子マーカー (変異/発現) の相関を調べるとき
+- DepMap CRISPR/RNAi 依存性データを併用するとき
+- 細胞株間の薬剤応答パターンを比較するとき
+- 新規化合物のスクリーニング結果を NCI-60 と比較するとき
+
+---
+
+## Quick Start
+
+## 1. CellMiner データ取得
+
+```python
+import requests
+import pandas as pd
+import numpy as np
+from io import StringIO
+
+CELLMINER_BASE = "https://discover.nci.nih.gov/cellminer/api"
+
+
+def cellminer_drug_activity(nsc_id=None, drug_name=None):
+    """
+    CellMiner — NCI-60 薬剤活性データ取得。
+
+    Parameters:
+        nsc_id: str — NSC ID (例: "740")
+        drug_name: str — 薬剤名 (例: "Paclitaxel")
+    """
+    if nsc_id:
+        url = f"{CELLMINER_BASE}/compound/{nsc_id}/activity"
+    elif drug_name:
+        url = f"{CELLMINER_BASE}/compound/search"
+        params = {"name": drug_name}
+        resp = requests.get(url, params=params, timeout=30)
+        resp.raise_for_status()
+        compounds = resp.json()
+        if not compounds:
+            print(f"Drug not found: {drug_name}")
+            return pd.DataFrame()
+        nsc_id = compounds[0].get("nsc", "")
+        url = f"{CELLMINER_BASE}/compound/{nsc_id}/activity"
+
+    resp = requests.get(url, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for cell_line, values in data.get("activity", {}).items():
+        results.append({
+            "cell_line": cell_line,
+            "tissue": values.get("tissue", ""),
+            "gi50_log": values.get("gi50", None),
+            "tgi_log": values.get("tgi", None),
+            "lc50_log": values.get("lc50", None),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"CellMiner: NSC {nsc_id} → {len(df)} cell lines")
+    return df
+```
+
+## 2. NCI-60 バルクデータ取得
+
+```python
+def nci60_bulk_download(data_type="drug_activity"):
+    """
+    NCI-60 バルクデータセット取得。
+
+    Parameters:
+        data_type: str — "drug_activity", "gene_expression",
+                         "mutation", "copy_number"
+    """
+    urls = {
+        "drug_activity": "https://discover.nci.nih.gov/cellminer/download/DTP_NCI60_ZSCORE.csv",
+        "gene_expression": "https://discover.nci.nih.gov/cellminer/download/GeneExpr_RMA.csv",
+        "mutation": "https://discover.nci.nih.gov/cellminer/download/Exome_Mutation.csv",
+    }
+
+    url = urls.get(data_type)
+    if not url:
+        raise ValueError(f"Unknown data type: {data_type}")
+
+    resp = requests.get(url, timeout=120)
+    resp.raise_for_status()
+
+    df = pd.read_csv(StringIO(resp.text))
+    print(f"NCI-60 bulk: {data_type} → {df.shape}")
+    return df
+```
+
+## 3. 薬剤-分子マーカー相関
+
+```python
+from scipy import stats
+
+
+def drug_marker_correlation(drug_activity, molecular_data,
+                             marker_type="expression", top_n=50):
+    """
+    薬剤感受性と分子マーカーの相関解析。
+
+    Parameters:
+        drug_activity: pd.DataFrame — GI50 データ (cell_line, gi50)
+        molecular_data: pd.DataFrame — 分子データ (cell_line, gene, value)
+        marker_type: str — "expression", "mutation", "copy_number"
+        top_n: int — 上位相関遺伝子数
+    """
+    # 細胞株一致
+    common_lines = set(drug_activity["cell_line"]) & set(molecular_data["cell_line"])
+    drug_sub = drug_activity[drug_activity["cell_line"].isin(common_lines)]
+    mol_sub = molecular_data[molecular_data["cell_line"].isin(common_lines)]
+
+    # 遺伝子ごとの相関
+    correlations = []
+    genes = mol_sub["gene"].unique() if "gene" in mol_sub.columns else mol_sub.columns[1:]
+
+    drug_values = drug_sub.set_index("cell_line")["gi50_log"]
+
+    for gene in genes:
+        if "gene" in mol_sub.columns:
+            gene_data = mol_sub[mol_sub["gene"] == gene].set_index("cell_line")["value"]
+        else:
+            gene_data = mol_sub.set_index("cell_line")[gene]
+
+        common = drug_values.index.intersection(gene_data.index)
+        if len(common) < 10:
+            continue
+
+        r, p = stats.pearsonr(drug_values[common], gene_data[common])
+        correlations.append({
+            "gene": gene,
+            "pearson_r": r,
+            "p_value": p,
+            "n_samples": len(common),
+        })
+
+    corr_df = pd.DataFrame(correlations)
+    corr_df["adj_p"] = corr_df["p_value"] * len(corr_df)  # Bonferroni
+    corr_df = corr_df.sort_values("p_value")
+
+    print(f"Drug-marker correlation: {len(corr_df)} genes tested, "
+          f"top |r| = {corr_df['pearson_r'].abs().max():.3f}")
+    return corr_df.head(top_n)
+```
+
+## 4. 組織別薬剤応答パターン
+
+```python
+def tissue_response_pattern(drug_activity, min_lines=3):
+    """
+    組織別の薬剤応答パターン解析。
+
+    Parameters:
+        drug_activity: pd.DataFrame — GI50 データ
+        min_lines: int — 最小細胞株数
+    """
+    tissue_stats = drug_activity.groupby("tissue").agg(
+        n_lines=("gi50_log", "count"),
+        mean_gi50=("gi50_log", "mean"),
+        std_gi50=("gi50_log", "std"),
+        min_gi50=("gi50_log", "min"),
+        max_gi50=("gi50_log", "max"),
+    ).reset_index()
+
+    tissue_stats = tissue_stats[tissue_stats["n_lines"] >= min_lines]
+    tissue_stats = tissue_stats.sort_values("mean_gi50")
+
+    # 感受性/耐性スコア
+    overall_mean = drug_activity["gi50_log"].mean()
+    tissue_stats["sensitivity_z"] = (
+        (tissue_stats["mean_gi50"] - overall_mean)
+        / drug_activity["gi50_log"].std()
+    )
+
+    print(f"Tissue patterns: {len(tissue_stats)} tissues")
+    for _, row in tissue_stats.iterrows():
+        label = "Sensitive" if row["sensitivity_z"] < -0.5 else (
+            "Resistant" if row["sensitivity_z"] > 0.5 else "Neutral"
+        )
+        print(f"  {row['tissue']}: GI50={row['mean_gi50']:.2f} ({label})")
+    return tissue_stats
+```
+
+## 5. DepMap 統合スクリーニング
+
+```python
+DEPMAP_BASE = "https://depmap.org/portal/api"
+
+
+def depmap_gene_dependency(gene_symbol, dataset="Chronos_Combined"):
+    """
+    DepMap — CRISPR/RNAi 遺伝子依存性取得。
+
+    Parameters:
+        gene_symbol: str — 遺伝子シンボル
+        dataset: str — データセット名
+    """
+    url = f"{DEPMAP_BASE}/download/custom"
+    params = {
+        "gene": gene_symbol,
+        "dataset": dataset,
+    }
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for entry in data.get("data", []):
+        results.append({
+            "cell_line": entry.get("cell_line_name", ""),
+            "lineage": entry.get("lineage", ""),
+            "dependency_score": entry.get("score", None),
+        })
+
+    df = pd.DataFrame(results)
+    if len(df) > 0:
+        n_dependent = (df["dependency_score"] < -0.5).sum()
+        print(f"DepMap {gene_symbol}: {len(df)} lines, "
+              f"{n_dependent} dependent (score < -0.5)")
+    return df
+```
+
+## 6. NCI-60 統合スクリーニングパイプライン
+
+```python
+def nci60_screening_pipeline(drug_name=None, nsc_id=None,
+                              target_gene=None, output_dir="results"):
+    """
+    NCI-60 + DepMap 統合スクリーニングパイプライン。
+
+    Parameters:
+        drug_name: str — 薬剤名
+        nsc_id: str — NSC ID
+        target_gene: str — 標的遺伝子 (DepMap 連携)
+        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) NCI-60 薬剤活性
+    drug_data = cellminer_drug_activity(nsc_id=nsc_id, drug_name=drug_name)
+    drug_data.to_csv(output_dir / "drug_activity.csv", index=False)
+
+    # 2) 組織別パターン
+    tissue_patterns = tissue_response_pattern(drug_data)
+    tissue_patterns.to_csv(output_dir / "tissue_patterns.csv", index=False)
+
+    # 3) 発現相関
+    expr_data = nci60_bulk_download("gene_expression")
+    correlations = drug_marker_correlation(drug_data, expr_data)
+    correlations.to_csv(output_dir / "marker_correlations.csv", index=False)
+
+    # 4) DepMap 連携 (標的遺伝子あれば)
+    if target_gene:
+        depmap_data = depmap_gene_dependency(target_gene)
+        depmap_data.to_csv(output_dir / "depmap_dependency.csv", index=False)
+
+    print(f"Pipeline complete: {output_dir}")
+    return {
+        "drug_activity": drug_data,
+        "tissue_patterns": tissue_patterns,
+        "correlations": correlations,
+    }
+```
+
+---
+
+## パイプライン統合
+
+```
+compound-screening → nci60-screening → precision-oncology
+  (ZINC/VS)           (NCI-60/DepMap)    (MTB レポート)
+       │                     │                 ↓
+drug-target-profiling ──────┘          cancer-genomics
+  (ChEMBL/DGIdb)        │              (COSMIC/DepMap)
+                         ↓
+                   cell-line-resources
+                   (Cellosaurus)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/drug_activity.csv` | NCI-60 GI50 データ | → precision-oncology |
+| `results/tissue_patterns.csv` | 組織別応答パターン | → cancer-genomics |
+| `results/marker_correlations.csv` | 薬剤-マーカー相関 | → drug-target-profiling |
+| `results/depmap_dependency.csv` | DepMap 依存性スコア | → cell-line-resources |

package/src/.github/skills/scientific-paleobiology/SKILL.md

@@ -0,0 +1,265 @@
+---
+name: scientific-paleobiology
+description: |
+  古生物学データベーススキル。Paleobiology Database (PBDB) REST
+  API による化石産出記録・分類群・コレクション検索、地質年代
+  多様性曲線・古地理解析。ToolUniverse 連携: paleobiology。
+tu_tools:
+  - key: paleobiology
+    name: Paleobiology Database
+    description: PBDB 化石産出記録・分類群・コレクション検索
+---
+
+# Scientific Paleobiology
+
+Paleobiology Database (PBDB) REST API を活用した古生物学的
+多様性解析パイプラインを提供する。
+
+## When to Use
+
+- 化石産出記録 (occurrence) を検索するとき
+- 分類群 (taxa) の地質年代分布を調べるとき
+- 化石コレクション/産地情報を検索するとき
+- 地質年代を通じた多様性曲線を作成するとき
+- 大量絶滅イベントのパターンを分析するとき
+- 古地理的分布を解析するとき
+
+---
+
+## Quick Start
+
+## 1. PBDB 化石産出記録検索
+
+```python
+import requests
+import pandas as pd
+import numpy as np
+
+PBDB_BASE = "https://paleobiodb.org/data1.2"
+
+
+def pbdb_search_occurrences(taxon=None, interval=None,
+                              lngmin=None, lngmax=None,
+                              latmin=None, latmax=None, limit=1000):
+    """
+    PBDB — 化石産出記録検索。
+
+    Parameters:
+        taxon: str — 分類群名 (例: "Dinosauria", "Trilobita")
+        interval: str — 地質年代区間 (例: "Cretaceous", "Permian")
+        lngmin: float — 経度最小値
+        lngmax: float — 経度最大値
+        latmin: float — 緯度最小値
+        latmax: float — 緯度最大値
+        limit: int — 最大結果数
+    """
+    url = f"{PBDB_BASE}/occs/list.json"
+    params = {
+        "show": "coords,phylo,time",
+        "limit": limit,
+    }
+    if taxon:
+        params["base_name"] = taxon
+    if interval:
+        params["interval"] = interval
+    if all(v is not None for v in [lngmin, lngmax, latmin, latmax]):
+        params.update({
+            "lngmin": lngmin, "lngmax": lngmax,
+            "latmin": latmin, "latmax": latmax,
+        })
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    records = resp.json().get("records", [])
+
+    results = []
+    for r in records:
+        results.append({
+            "occurrence_no": r.get("oid", ""),
+            "taxon_name": r.get("tna", ""),
+            "taxon_rank": r.get("rnk", ""),
+            "phylum": r.get("phl", ""),
+            "class": r.get("cll", ""),
+            "order": r.get("odl", ""),
+            "family": r.get("fml", ""),
+            "early_interval": r.get("oei", ""),
+            "late_interval": r.get("oli", ""),
+            "max_ma": r.get("eag", None),
+            "min_ma": r.get("lag", None),
+            "lng": r.get("lng", None),
+            "lat": r.get("lat", None),
+            "collection_no": r.get("cid", ""),
+            "reference_no": r.get("rid", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PBDB occurrences: {len(df)} records "
+          f"(taxon={taxon}, interval={interval})")
+    return df
+```
+
+## 2. PBDB 分類群情報検索
+
+```python
+def pbdb_search_taxa(name=None, rank=None, interval=None, limit=500):
+    """
+    PBDB — 分類群検索。
+
+    Parameters:
+        name: str — 分類群名 (例: "Dinosauria")
+        rank: str — ランク (例: "genus", "family", "order")
+        interval: str — 地質年代区間
+        limit: int — 最大結果数
+    """
+    url = f"{PBDB_BASE}/taxa/list.json"
+    params = {
+        "show": "attr,app,size",
+        "limit": limit,
+    }
+    if name:
+        params["base_name"] = name
+    if rank:
+        params["rank"] = rank
+    if interval:
+        params["interval"] = interval
+
+    resp = requests.get(url, params=params, timeout=30)
+    resp.raise_for_status()
+    records = resp.json().get("records", [])
+
+    results = []
+    for r in records:
+        results.append({
+            "taxon_no": r.get("oid", ""),
+            "taxon_name": r.get("nam", ""),
+            "rank": r.get("rnk", ""),
+            "parent_name": r.get("prl", ""),
+            "n_occs": r.get("noc", 0),
+            "first_appearance": r.get("fea", ""),
+            "last_appearance": r.get("lla", ""),
+            "extant": r.get("ext", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PBDB taxa: {len(df)} records (name={name})")
+    return df
+```
+
+## 3. 地質年代多様性曲線
+
+```python
+def pbdb_diversity_curve(taxon, time_resolution="stage",
+                           rank="genus"):
+    """
+    PBDB — 地質年代多様性曲線生成。
+
+    Parameters:
+        taxon: str — 分類群名
+        time_resolution: str — "stage" or "epoch" or "period"
+        rank: str — カウントするランク ("genus", "family")
+    """
+    url = f"{PBDB_BASE}/occs/diversity.json"
+    params = {
+        "base_name": taxon,
+        "count": rank,
+        "time_reso": time_resolution,
+    }
+    resp = requests.get(url, params=params, timeout=60)
+    resp.raise_for_status()
+    records = resp.json().get("records", [])
+
+    results = []
+    for r in records:
+        results.append({
+            "interval_name": r.get("idn", ""),
+            "max_ma": r.get("eag", None),
+            "min_ma": r.get("lag", None),
+            "mid_ma": (float(r.get("eag", 0)) +
+                       float(r.get("lag", 0))) / 2,
+            "sampled_in_bin": r.get("dsb", 0),
+            "n_originations": r.get("dor", 0),
+            "n_extinctions": r.get("dex", 0),
+            "range_through": r.get("drt", 0),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"PBDB diversity: {len(df)} intervals, "
+          f"max diversity={df['sampled_in_bin'].max()} {rank}")
+    return df
+```
+
+## 4. 古生物学統合パイプライン
+
+```python
+def paleobiology_pipeline(taxon, interval=None,
+                            output_dir="results"):
+    """
+    古生物学統合パイプライン。
+
+    Parameters:
+        taxon: str — 分類群名 (例: "Dinosauria")
+        interval: str — 地質年代区間 (オプション)
+        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) 産出記録
+    occ = pbdb_search_occurrences(taxon=taxon, interval=interval)
+    occ.to_csv(output_dir / "occurrences.csv", index=False)
+
+    # 2) 分類群情報
+    taxa = pbdb_search_taxa(name=taxon)
+    taxa.to_csv(output_dir / "taxa.csv", index=False)
+
+    # 3) 多様性曲線
+    diversity = pbdb_diversity_curve(taxon)
+    diversity.to_csv(output_dir / "diversity.csv", index=False)
+
+    # 4) 地理的サマリ
+    if "lat" in occ.columns and "lng" in occ.columns:
+        geo_summary = occ.groupby("early_interval").agg(
+            n_records=("occurrence_no", "count"),
+            mean_lat=("lat", "mean"),
+            mean_lng=("lng", "mean"),
+        ).reset_index()
+        geo_summary.to_csv(output_dir / "geo_summary.csv", index=False)
+
+    print(f"Paleobiology pipeline: {output_dir}")
+    return {
+        "occurrences": occ,
+        "taxa": taxa,
+        "diversity": diversity,
+    }
+```
+
+---
+
+## ToolUniverse 連携
+
+| TU Key | ツール名 | 連携内容 |
+|--------|---------|---------|
+| `paleobiology` | Paleobiology Database | 化石産出・分類群・コレクション検索 |
+
+## パイプライン統合
+
+```
+phylogenetics → paleobiology → environmental-ecology
+  (系統解析)   (化石記録)      (GBIF/生態)
+      │              │                ↓
+  taxonomy ─────────┘         environmental-geodata
+  (分類体系)    │              (環境モデリング)
+                ↓
+         macroevolution
+         (大進化パターン)
+```
+
+## パイプライン出力
+
+| ファイル | 説明 | 次スキル |
+|---------|------|---------|
+| `results/occurrences.csv` | 化石産出記録 | → environmental-ecology |
+| `results/taxa.csv` | 分類群情報 | → phylogenetics |
+| `results/diversity.csv` | 多様性曲線 | → macroevolution |
+| `results/geo_summary.csv` | 古地理サマリ | → environmental-geodata |