@nahisaho/satori 0.11.1 → 0.12.0

package/src/.github/skills/scientific-protein-interaction-network/SKILL.md

@@ -0,0 +1,352 @@
+---
+name: scientific-protein-interaction-network
+description: |
+  タンパク質-タンパク質相互作用 (PPI) ネットワーク解析スキル。STRING、IntAct、
+  BioGRID、STITCH (化学-タンパク質) 相互作用データベースを統合した
+  ネットワーク構築・解析パイプライン。GO/KEGG 富化、相互作用パートナー発見、
+  組織特異的ネットワーク (HumanBase)、化合物-標的ネットワーク対応。
+  14 の ToolUniverse SMCP ツールと連携。
+---
+
+# Scientific Protein Interaction Network
+
+STRING / IntAct / BioGRID / STITCH の 4 大 PPI データベースを統合した
+タンパク質相互作用ネットワーク解析パイプラインを提供する。
+
+## When to Use
+
+- DEG や変異遺伝子の PPI ネットワークを構築するとき
+- ハブタンパク質やボトルネックの特定が必要なとき
+- 化合物と標的タンパク質の相互作用を調べるとき
+- 組織特異的な相互作用ネットワークを評価するとき
+- PPI データに基づく GO/KEGG 富化やモジュール解析を行うとき
+
+---
+
+## Quick Start
+
+## 1. STRING PPI ネットワーク取得
+
+```python
+import requests
+import pandas as pd
+import networkx as nx
+
+
+def string_get_interactions(proteins, species=9606,
+                             score_threshold=400,
+                             network_type="functional"):
+    """
+    STRING API v12 による PPI ネットワーク取得。
+
+    Parameters:
+        proteins: list — タンパク質名/UniProt ID リスト
+        species: int — NCBI Taxonomy ID (9606=Homo sapiens)
+        score_threshold: int — 信頼スコア閾値 (0-1000)
+        network_type: "functional" or "physical"
+    """
+    base = "https://string-db.org/api/json"
+
+    # タンパク質 ID 解決
+    resolve_url = f"{base}/get_string_ids"
+    resolved = []
+    for batch in [proteins[i:i+10] for i in range(0, len(proteins), 10)]:
+        params = {
+            "identifiers": "\r".join(batch),
+            "species": species,
+            "limit": 1,
+        }
+        resp = requests.get(resolve_url, params=params)
+        for r in resp.json():
+            resolved.append(r["stringId"])
+
+    if not resolved:
+        print("No proteins resolved")
+        return pd.DataFrame(), nx.Graph()
+
+    # 相互作用取得
+    interaction_url = f"{base}/network"
+    params = {
+        "identifiers": "\r".join(resolved),
+        "species": species,
+        "required_score": score_threshold,
+        "network_type": network_type,
+    }
+    resp = requests.get(interaction_url, params=params)
+    interactions = resp.json()
+
+    edges = []
+    for i in interactions:
+        edges.append({
+            "protein_a": i["preferredName_A"],
+            "protein_b": i["preferredName_B"],
+            "score": i["score"],
+            "nscore": i.get("nscore", 0),
+            "fscore": i.get("fscore", 0),
+            "pscore": i.get("pscore", 0),
+            "ascore": i.get("ascore", 0),
+            "escore": i.get("escore", 0),
+            "dscore": i.get("dscore", 0),
+            "tscore": i.get("tscore", 0),
+        })
+
+    df = pd.DataFrame(edges)
+
+    # NetworkX グラフ構築
+    G = nx.Graph()
+    for _, row in df.iterrows():
+        G.add_edge(row["protein_a"], row["protein_b"],
+                   weight=row["score"] / 1000.0)
+
+    print(f"STRING network: {G.number_of_nodes()} nodes, "
+          f"{G.number_of_edges()} edges (score ≥ {score_threshold})")
+    return df, G
+```
+
+## 2. IntAct 分子相互作用検索
+
+```python
+def intact_search_interactions(query, species="human",
+                                interaction_type=None,
+                                max_results=200):
+    """
+    IntAct REST API による分子相互作用検索。
+
+    Parameters:
+        query: str — タンパク質名/UniProt ID
+        species: str or int — "human" or taxonomy ID
+        interaction_type: str — MI term (e.g., "MI:0407" physical association)
+    """
+    url = "https://www.ebi.ac.uk/intact/ws/interaction/findInteractions"
+    params = {
+        "query": query,
+        "maxResults": max_results,
+    }
+    if species:
+        params["species"] = species
+
+    resp = requests.get(url, params=params)
+    if resp.status_code != 200:
+        print(f"IntAct error: {resp.status_code}")
+        return pd.DataFrame()
+
+    data = resp.json()
+    interactions = data.get("content", [])
+
+    results = []
+    for ix in interactions:
+        interactor_a = ix.get("interactorA", {})
+        interactor_b = ix.get("interactorB", {})
+        results.append({
+            "interactor_a": interactor_a.get("preferredIdentifier", ""),
+            "interactor_a_name": interactor_a.get("shortLabel", ""),
+            "interactor_b": interactor_b.get("preferredIdentifier", ""),
+            "interactor_b_name": interactor_b.get("shortLabel", ""),
+            "interaction_type": ix.get("interactionType", ""),
+            "detection_method": ix.get("detectionMethod", ""),
+            "confidence": ix.get("confidenceValue", 0),
+            "publication": ix.get("pubmedId", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"IntAct: {len(df)} interactions for '{query}'")
+    return df
+```
+
+## 3. STITCH 化合物-タンパク質相互作用
+
+```python
+def stitch_chemical_protein(chemicals, species=9606,
+                             score_threshold=400):
+    """
+    STITCH API による化合物-タンパク質相互作用検索。
+
+    Parameters:
+        chemicals: list — 化合物名/CID リスト
+        species: int — NCBI Taxonomy ID
+        score_threshold: int — 信頼スコア閾値
+    """
+    url = "http://stitch.embl.de/api/json/interactionsList"
+    params = {
+        "identifiers": "\r".join(chemicals),
+        "species": species,
+        "required_score": score_threshold,
+    }
+
+    resp = requests.get(url, params=params)
+    interactions = resp.json()
+
+    results = []
+    for i in interactions:
+        results.append({
+            "chemical": i.get("preferredName_A", ""),
+            "protein": i.get("preferredName_B", ""),
+            "score": i.get("score", 0),
+            "type_a": "chemical" if i.get("ncbiTaxonId_A") == -1 else "protein",
+        })
+
+    df = pd.DataFrame(results)
+    print(f"STITCH: {len(df)} chemical-protein interactions")
+    return df
+```
+
+## 4. PPI ネットワーク解析 (中心性・モジュール)
+
+```python
+def analyze_ppi_network(G, community_method="louvain"):
+    """
+    PPI ネットワークのトポロジー解析。
+
+    Parameters:
+        G: nx.Graph — PPI ネットワーク
+        community_method: "louvain" or "label_propagation"
+    """
+    if G.number_of_nodes() == 0:
+        return {}
+
+    # 中心性指標
+    degree_cent = nx.degree_centrality(G)
+    betweenness = nx.betweenness_centrality(G)
+    closeness = nx.closeness_centrality(G)
+
+    # ハブタンパク質 (degree top 10)
+    hubs = sorted(degree_cent.items(), key=lambda x: -x[1])[:10]
+
+    # ボトルネック (betweenness top 10)
+    bottlenecks = sorted(betweenness.items(), key=lambda x: -x[1])[:10]
+
+    # コミュニティ検出
+    if community_method == "louvain":
+        from community import community_louvain
+        partition = community_louvain.best_partition(G)
+    else:
+        communities = nx.community.label_propagation_communities(G)
+        partition = {}
+        for i, comm in enumerate(communities):
+            for node in comm:
+                partition[node] = i
+
+    n_communities = len(set(partition.values()))
+
+    stats = {
+        "nodes": G.number_of_nodes(),
+        "edges": G.number_of_edges(),
+        "density": round(nx.density(G), 4),
+        "avg_clustering": round(nx.average_clustering(G), 4),
+        "connected_components": nx.number_connected_components(G),
+        "communities": n_communities,
+        "hub_proteins": [h[0] for h in hubs],
+        "bottleneck_proteins": [b[0] for b in bottlenecks],
+    }
+
+    centrality_df = pd.DataFrame({
+        "protein": list(degree_cent.keys()),
+        "degree_centrality": list(degree_cent.values()),
+        "betweenness": [betweenness[n] for n in degree_cent.keys()],
+        "closeness": [closeness[n] for n in degree_cent.keys()],
+        "community": [partition.get(n, -1) for n in degree_cent.keys()],
+    }).sort_values("degree_centrality", ascending=False)
+
+    print(f"PPI analysis: {stats['nodes']} nodes, {stats['edges']} edges, "
+          f"{n_communities} communities")
+    return stats, centrality_df, partition
+```
+
+## 5. PPI ネットワーク可視化
+
+```python
+def visualize_ppi_network(G, partition=None, hub_proteins=None,
+                           output="figures/ppi_network.png",
+                           layout="spring"):
+    """
+    PPI ネットワークの可視化。
+    """
+    import matplotlib.pyplot as plt
+    import os
+    os.makedirs(os.path.dirname(output), exist_ok=True)
+
+    fig, ax = plt.subplots(figsize=(14, 14))
+
+    if layout == "spring":
+        pos = nx.spring_layout(G, k=1.5, seed=42)
+    elif layout == "kamada_kawai":
+        pos = nx.kamada_kawai_layout(G)
+
+    # ノードサイズ = 次数
+    node_sizes = [300 + 100 * G.degree(n) for n in G.nodes()]
+
+    # コミュニティカラー
+    if partition:
+        import matplotlib.cm as cm
+        n_comm = len(set(partition.values()))
+        colors = [cm.Set3(partition.get(n, 0) / max(n_comm, 1)) for n in G.nodes()]
+    else:
+        colors = "steelblue"
+
+    nx.draw_networkx_edges(G, pos, alpha=0.2, ax=ax)
+    nx.draw_networkx_nodes(G, pos, node_size=node_sizes,
+                           node_color=colors, alpha=0.8, ax=ax)
+
+    # ハブタンパク質のラベル
+    if hub_proteins:
+        labels = {n: n for n in G.nodes() if n in hub_proteins}
+    else:
+        labels = {n: n for n in G.nodes() if G.degree(n) >= 5}
+    nx.draw_networkx_labels(G, pos, labels, font_size=8, ax=ax)
+
+    ax.set_title(f"PPI Network ({G.number_of_nodes()} proteins, "
+                 f"{G.number_of_edges()} interactions)")
+    ax.axis("off")
+    plt.tight_layout()
+    plt.savefig(output, dpi=300, bbox_inches="tight")
+    plt.close()
+    print(f"Saved: {output}")
+```
+
+## References
+
+### Output Files
+
+| ファイル | 形式 |
+|---|---|
+| `results/string_interactions.csv` | CSV |
+| `results/intact_interactions.csv` | CSV |
+| `results/stitch_interactions.csv` | CSV |
+| `results/ppi_centrality.csv` | CSV |
+| `results/ppi_network.graphml` | GraphML |
+| `figures/ppi_network.png` | PNG |
+
+### 利用可能ツール
+
+> [ToolUniverse](https://github.com/mims-harvard/ToolUniverse) SMCP 経由で利用可能な外部ツール。
+
+| カテゴリ | 主要ツール | 用途 |
+|---|---|---|
+| IntAct | `intact_search_interactions` | 分子相互作用検索 |
+| IntAct | `intact_get_interactions` | 相互作用データ取得 |
+| IntAct | `intact_get_interactor` | 相互作用因子詳細 |
+| IntAct | `intact_get_interaction_details` | 相互作用詳細 |
+| IntAct | `intact_get_interaction_network` | ネットワーク取得 |
+| IntAct | `intact_get_interactions_by_organism` | 生物種別相互作用 |
+| IntAct | `intact_get_interactions_by_complex` | 複合体別相互作用 |
+| IntAct | `intact_get_complex_details` | 複合体詳細 |
+| STRING/BioGRID | `STRING_get_protein_interactions` | STRING PPI 取得 |
+| STRING/BioGRID | `BioGRID_get_interactions` | BioGRID 相互作用取得 |
+| STITCH | `STITCH_get_chemical_protein_interactions` | 化合物-タンパク質相互作用 |
+| STITCH | `STITCH_get_interaction_partners` | 相互作用パートナー |
+| STITCH | `STITCH_resolve_identifier` | 化合物 ID 解決 |
+| HumanBase | `humanbase_ppi_analysis` | 組織特異的 PPI 解析 |
+
+### 参照スキル
+
+| スキル | 関連 |
+|---|---|
+| `scientific-drug-target-profiling` | 標的タンパク質 → PPI 拡張 |
+| `scientific-network-analysis` | 汎用ネットワーク解析手法 |
+| `scientific-pathway-enrichment` | PPI モジュール → パスウェイ富化 |
+| `scientific-protein-structure-analysis` | 構造情報 → 相互作用界面 |
+| `scientific-systems-biology` | GRN ↔ PPI 統合 |
+
+### 依存パッケージ
+
+`networkx`, `requests`, `pandas`, `matplotlib`, `python-louvain` (community)