@nahisaho/satori 0.11.1 → 0.13.0

package/src/.github/skills/scientific-clinical-reporting/SKILL.md

@@ -0,0 +1,324 @@
+---
+name: scientific-clinical-reporting
+description: |
+  臨床レポート自動生成スキル。検査結果サマリー (SOAP ノート)、バイオマーカー
+  プロファイルレポート、薬理ゲノミクスレポート、臨床試験要約を構造化テンプレート
+  (PDF/LaTeX/HTML) で出力。HL7 FHIR DiagnosticReport 形式にも対応。
+---
+
+# Scientific Clinical Reporting
+
+臨床データから構造化レポートを自動生成するパイプラインを提供する。
+
+## When to Use
+
+- 検査結果を SOAP ノート形式でまとめるとき
+- バイオマーカープロファイルレポートを作成するとき
+- ファーマコゲノミクスレポート (CPIC ガイドライン準拠) が必要なとき
+- 臨床試験の CSR (Clinical Study Report) サマリーを生成するとき
+- HL7 FHIR DiagnosticReport 形式で出力するとき
+
+---
+
+## Quick Start
+
+## 1. SOAP ノート生成
+
+```python
+import json
+from datetime import datetime
+
+
+def generate_soap_note(patient_data, findings, assessment, plan):
+    """
+    SOAP ノート形式の臨床レポートを生成。
+
+    Parameters:
+        patient_data: dict — {"id": "...", "age": 45, "sex": "M", ...}
+        findings: dict — {"subjective": [...], "objective": [...]}
+        assessment: list — 評価・診断リスト
+        plan: list — 治療計画リスト
+    """
+    soap = {
+        "report_type": "SOAP_Note",
+        "generated_at": datetime.now().isoformat(),
+        "patient": {
+            "id": patient_data.get("id", "ANON"),
+            "age": patient_data.get("age"),
+            "sex": patient_data.get("sex"),
+        },
+        "S": {  # Subjective
+            "chief_complaint": findings.get("chief_complaint", ""),
+            "history": findings.get("subjective", []),
+        },
+        "O": {  # Objective
+            "vitals": findings.get("vitals", {}),
+            "lab_results": findings.get("lab_results", []),
+            "imaging": findings.get("imaging", []),
+            "physical_exam": findings.get("objective", []),
+        },
+        "A": {  # Assessment
+            "diagnoses": assessment,
+            "differential": findings.get("differential", []),
+        },
+        "P": {  # Plan
+            "treatment": plan,
+            "follow_up": findings.get("follow_up", ""),
+            "referrals": findings.get("referrals", []),
+        },
+    }
+
+    print(f"SOAP note: patient={soap['patient']['id']}, "
+          f"diagnoses={len(assessment)}, plans={len(plan)}")
+    return soap
+```
+
+## 2. バイオマーカープロファイルレポート
+
+```python
+import pandas as pd
+
+
+def biomarker_profile_report(biomarkers_df, reference_ranges=None):
+    """
+    バイオマーカープロファイルレポート生成。
+
+    Parameters:
+        biomarkers_df: DataFrame — columns: [marker, value, unit, specimen]
+        reference_ranges: dict — {"marker": {"low": x, "high": y, "unit": "..."}}
+    """
+    if reference_ranges is None:
+        reference_ranges = {
+            "CEA": {"low": 0, "high": 5.0, "unit": "ng/mL"},
+            "AFP": {"low": 0, "high": 10.0, "unit": "ng/mL"},
+            "CA19-9": {"low": 0, "high": 37.0, "unit": "U/mL"},
+            "CA125": {"low": 0, "high": 35.0, "unit": "U/mL"},
+            "PSA": {"low": 0, "high": 4.0, "unit": "ng/mL"},
+            "HER2": {"low": 0, "high": 1, "unit": "IHC score"},
+            "Ki-67": {"low": 0, "high": 14, "unit": "%"},
+            "PD-L1 TPS": {"low": 0, "high": 1, "unit": "%"},
+        }
+
+    results = []
+    for _, row in biomarkers_df.iterrows():
+        marker = row["marker"]
+        value = float(row["value"])
+        ref = reference_ranges.get(marker, {})
+
+        status = "normal"
+        if ref:
+            if value > ref.get("high", float("inf")):
+                status = "HIGH"
+            elif value < ref.get("low", float("-inf")):
+                status = "LOW"
+
+        results.append({
+            "marker": marker,
+            "value": value,
+            "unit": row.get("unit", ref.get("unit", "")),
+            "reference": f"{ref.get('low', '?')}-{ref.get('high', '?')}",
+            "status": status,
+        })
+
+    report_df = pd.DataFrame(results)
+    abnormal = report_df[report_df["status"] != "normal"]
+
+    report = {
+        "report_type": "Biomarker_Profile",
+        "total_markers": len(report_df),
+        "abnormal_count": len(abnormal),
+        "results": report_df.to_dict("records"),
+        "summary": (
+            f"{len(abnormal)} of {len(report_df)} markers outside reference range"
+            if len(abnormal) > 0
+            else "All markers within reference range"
+        ),
+    }
+
+    print(f"Biomarker profile: {len(report_df)} markers, "
+          f"{len(abnormal)} abnormal")
+    return report
+```
+
+## 3. ファーマコゲノミクスレポート
+
+```python
+def pharmacogenomics_report(genotypes, medications):
+    """
+    CPIC ガイドライン準拠のファーマコゲノミクスレポート。
+
+    Parameters:
+        genotypes: dict — {"CYP2D6": "*1/*4", "CYP2C19": "*1/*2", ...}
+        medications: list — ["codeine", "clopidogrel", ...]
+    """
+    # CPIC phenotype マッピング (簡略)
+    cpic_phenotypes = {
+        "CYP2D6": {
+            "*1/*1": "Normal Metabolizer",
+            "*1/*4": "Intermediate Metabolizer",
+            "*4/*4": "Poor Metabolizer",
+            "*1/*2xN": "Ultrarapid Metabolizer",
+        },
+        "CYP2C19": {
+            "*1/*1": "Normal Metabolizer",
+            "*1/*2": "Intermediate Metabolizer",
+            "*2/*2": "Poor Metabolizer",
+            "*1/*17": "Rapid Metabolizer",
+            "*17/*17": "Ultrarapid Metabolizer",
+        },
+    }
+
+    # 推奨アクション (簡略)
+    drug_gene_map = {
+        "codeine": {"gene": "CYP2D6", "action": {
+            "Poor Metabolizer": "AVOID — use alternative analgesic",
+            "Ultrarapid Metabolizer": "AVOID — toxicity risk",
+            "Intermediate Metabolizer": "Use with caution, consider alternative",
+        }},
+        "clopidogrel": {"gene": "CYP2C19", "action": {
+            "Poor Metabolizer": "Use alternative antiplatelet (e.g., prasugrel)",
+            "Intermediate Metabolizer": "Consider alternative antiplatelet",
+        }},
+    }
+
+    recommendations = []
+    for drug in medications:
+        entry = drug_gene_map.get(drug, {})
+        gene = entry.get("gene", "Unknown")
+        genotype = genotypes.get(gene, "Unknown")
+        phenotype_map = cpic_phenotypes.get(gene, {})
+        phenotype = phenotype_map.get(genotype, "Indeterminate")
+
+        action = entry.get("action", {}).get(phenotype, "Standard dosing")
+        recommendations.append({
+            "drug": drug,
+            "gene": gene,
+            "genotype": genotype,
+            "phenotype": phenotype,
+            "recommendation": action,
+            "cpic_level": "A" if drug in drug_gene_map else "N/A",
+        })
+
+    report = {
+        "report_type": "Pharmacogenomics",
+        "genotypes_tested": genotypes,
+        "medications_queried": medications,
+        "recommendations": recommendations,
+    }
+
+    print(f"PGx report: {len(genotypes)} genes, {len(medications)} drugs, "
+          f"{len(recommendations)} recommendations")
+    return report
+```
+
+## 4. 構造化レポート出力 (LaTeX/HTML)
+
+```python
+def export_clinical_report(report, output_format="html",
+                            output_path="reports/clinical_report"):
+    """
+    臨床レポートを LaTeX/HTML/FHIR JSON 形式で出力。
+
+    Parameters:
+        report: dict — SOAP, Biomarker, PGx レポート
+        output_format: "html", "latex", "fhir_json"
+        output_path: str — 出力先パス (拡張子なし)
+    """
+    import os
+    os.makedirs(os.path.dirname(output_path), exist_ok=True)
+    report_type = report.get("report_type", "Clinical")
+
+    if output_format == "html":
+        filepath = f"{output_path}.html"
+        html_parts = [
+            "<!DOCTYPE html><html><head>",
+            f"<title>{report_type} Report</title>",
+            "<style>body{font-family:Arial;margin:2em;}"
+            "table{border-collapse:collapse;width:100%;}"
+            "td,th{border:1px solid #ddd;padding:8px;}</style>",
+            "</head><body>",
+            f"<h1>{report_type} Report</h1>",
+        ]
+
+        if report_type == "SOAP_Note":
+            for section in ["S", "O", "A", "P"]:
+                html_parts.append(f"<h2>{section}</h2>")
+                html_parts.append(f"<pre>{json.dumps(report.get(section, {}), indent=2, ensure_ascii=False)}</pre>")
+
+        elif report_type == "Biomarker_Profile":
+            html_parts.append("<table><tr><th>Marker</th><th>Value</th>"
+                              "<th>Reference</th><th>Status</th></tr>")
+            for r in report.get("results", []):
+                status_color = "red" if r["status"] != "normal" else "green"
+                html_parts.append(
+                    f"<tr><td>{r['marker']}</td><td>{r['value']} {r['unit']}</td>"
+                    f"<td>{r['reference']}</td>"
+                    f"<td style='color:{status_color}'>{r['status']}</td></tr>"
+                )
+            html_parts.append("</table>")
+
+        html_parts.append("</body></html>")
+        with open(filepath, "w") as f:
+            f.write("\n".join(html_parts))
+
+    elif output_format == "fhir_json":
+        filepath = f"{output_path}.fhir.json"
+        fhir = {
+            "resourceType": "DiagnosticReport",
+            "status": "final",
+            "category": [{"coding": [{"system": "http://terminology.hl7.org/CodeSystem/v2-0074",
+                                       "code": "LAB"}]}],
+            "code": {"text": report_type},
+            "issued": report.get("generated_at", datetime.now().isoformat()),
+            "result": [],
+        }
+        with open(filepath, "w") as f:
+            json.dump(fhir, f, indent=2)
+
+    elif output_format == "latex":
+        filepath = f"{output_path}.tex"
+        with open(filepath, "w") as f:
+            f.write(f"\\documentclass{{article}}\n")
+            f.write(f"\\title{{{report_type} Report}}\n")
+            f.write("\\begin{document}\n\\maketitle\n")
+            f.write(f"Report type: {report_type}\n")
+            f.write("\\end{document}\n")
+
+    print(f"Report exported: {filepath}")
+    return filepath
+```
+
+## References
+
+### Output Files
+
+| ファイル | 形式 |
+|---|---|
+| `reports/soap_note.json` | JSON |
+| `reports/biomarker_profile.json` | JSON |
+| `reports/pgx_report.json` | JSON |
+| `reports/clinical_report.html` | HTML |
+| `reports/clinical_report.tex` | LaTeX |
+| `reports/clinical_report.fhir.json` | FHIR JSON |
+
+### 利用可能ツール
+
+> 本スキルは ToolUniverse ツールに直接依存しない。
+
+| カテゴリ | 主要ツール | 用途 |
+|---|---|---|
+| — | — | — |
+
+### 参照スキル
+
+| スキル | 関連 |
+|---|---|
+| `scientific-variant-interpretation` | バリアント解釈レポート |
+| `scientific-variant-effect-prediction` | バリアント病原性スコア |
+| `scientific-pharmacogenomics` | PGx ガイドライン |
+| `scientific-precision-oncology` | 精密腫瘍学レポート |
+| `scientific-disease-research` | 疾患情報統合 |
+
+### 依存パッケージ
+
+`pandas`, `json` (stdlib), `datetime` (stdlib)

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

@@ -0,0 +1,245 @@
+---
+name: scientific-compound-screening
+description: |
+  化合物スクリーニングスキル。ZINC データベースを活用した購入可能化合物検索、
+  SMILES/名前ベースの類似性検索、カタログフィルタリング、
+  バーチャルスクリーニング前処理パイプライン。
+---
+
+# Scientific Compound Screening
+
+ZINC データベースを活用した化合物ライブラリ検索・
+バーチャルスクリーニング前処理パイプラインを提供する。
+
+## When to Use
+
+- 購入可能な化合物ライブラリを検索するとき
+- SMILES 構造式から類似化合物を探すとき
+- 化合物名からデータベースレコードを取得するとき
+- ベンダーカタログの絞り込みを行うとき
+- バーチャルスクリーニング用の化合物セットを準備するとき
+
+---
+
+## Quick Start
+
+## 1. ZINC 化合物名検索
+
+```python
+import requests
+import pandas as pd
+
+ZINC_API = "https://zinc15.docking.org"
+
+
+def zinc_search_by_name(name, max_results=20):
+    """
+    ZINC データベースで化合物名による検索。
+
+    Parameters:
+        name: str — compound name (e.g., "aspirin")
+        max_results: int — maximum results
+
+    ToolUniverse:
+        ZINC_search_by_name(name=name)
+    """
+    url = f"{ZINC_API}/substances/search"
+    params = {"q": name, "count": max_results}
+    resp = requests.get(url, params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data:
+        results.append({
+            "zinc_id": item.get("zinc_id", ""),
+            "name": item.get("name", ""),
+            "smiles": item.get("smiles", ""),
+            "mwt": item.get("mwt", ""),
+            "logp": item.get("logp", ""),
+            "purchasable": item.get("purchasability", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"ZINC search '{name}': {len(df)} compounds")
+    return df
+```
+
+## 2. ZINC SMILES 類似性検索
+
+```python
+def zinc_search_by_smiles(smiles, similarity=0.7, max_results=20):
+    """
+    ZINC で SMILES 構造式による類似性検索。
+
+    Parameters:
+        smiles: str — SMILES string
+        similarity: float — Tanimoto similarity threshold (0-1)
+
+    ToolUniverse:
+        ZINC_search_by_smiles(smiles=smiles)
+    """
+    url = f"{ZINC_API}/substances/search"
+    params = {
+        "smiles": smiles,
+        "similarity": similarity,
+        "count": max_results,
+    }
+    resp = requests.get(url, params=params)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for item in data:
+        results.append({
+            "zinc_id": item.get("zinc_id", ""),
+            "smiles": item.get("smiles", ""),
+            "similarity": item.get("similarity", ""),
+            "mwt": item.get("mwt", ""),
+            "logp": item.get("logp", ""),
+            "purchasable": item.get("purchasability", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"ZINC SMILES search: {len(df)} similar compounds "
+          f"(threshold={similarity})")
+    return df
+```
+
+## 3. ZINC 化合物詳細取得
+
+```python
+def zinc_get_substance(zinc_id):
+    """
+    ZINC ID から化合物の完全情報を取得。
+
+    Parameters:
+        zinc_id: str — ZINC ID (e.g., "ZINC000000000001")
+
+    ToolUniverse:
+        ZINC_get_substance(zinc_id=zinc_id)
+    """
+    url = f"{ZINC_API}/substances/{zinc_id}.json"
+    resp = requests.get(url)
+    resp.raise_for_status()
+    data = resp.json()
+
+    info = {
+        "zinc_id": data.get("zinc_id", ""),
+        "name": data.get("name", ""),
+        "smiles": data.get("smiles", ""),
+        "inchikey": data.get("inchikey", ""),
+        "mwt": data.get("mwt", ""),
+        "logp": data.get("logp", ""),
+        "num_rotatable_bonds": data.get("num_rotatable_bonds", ""),
+        "num_hba": data.get("num_hba", ""),
+        "num_hbd": data.get("num_hbd", ""),
+        "tpsa": data.get("tpsa", ""),
+        "purchasable": data.get("purchasability", ""),
+    }
+
+    print(f"ZINC {zinc_id}: {info['name']} (MW={info['mwt']})")
+    return info, data
+```
+
+## 4. ZINC カタログ一覧
+
+```python
+def zinc_get_catalogs():
+    """
+    ZINC の利用可能カタログ (ベンダー) 一覧を取得。
+
+    ToolUniverse:
+        ZINC_get_catalogs()
+    """
+    url = f"{ZINC_API}/catalogs.json"
+    resp = requests.get(url)
+    resp.raise_for_status()
+    data = resp.json()
+
+    results = []
+    for cat in data:
+        results.append({
+            "catalog_name": cat.get("name", ""),
+            "short_name": cat.get("short_name", ""),
+            "num_substances": cat.get("num_substances", 0),
+            "url": cat.get("url", ""),
+        })
+
+    df = pd.DataFrame(results)
+    print(f"ZINC catalogs: {len(df)} vendors")
+    return df
+```
+
+## 5. バーチャルスクリーニング前処理パイプライン
+
+```python
+def virtual_screening_prep(query_smiles, lipinski=True, max_compounds=100):
+    """
+    バーチャルスクリーニング用の化合物セット準備。
+    Lipinski's Rule of Five フィルタリング含む。
+
+    ToolUniverse (横断):
+        ZINC_search_by_smiles(smiles=query_smiles) → ZINC_get_substance(zinc_id)
+    """
+    # Step 1: Similar compound search
+    df = zinc_search_by_smiles(query_smiles, similarity=0.6,
+                               max_results=max_compounds)
+
+    if df.empty:
+        print("No similar compounds found")
+        return df
+
+    # Step 2: Lipinski filter
+    if lipinski:
+        df["mwt"] = pd.to_numeric(df["mwt"], errors="coerce")
+        df["logp"] = pd.to_numeric(df["logp"], errors="coerce")
+        before = len(df)
+        df = df[
+            (df["mwt"] <= 500)
+            & (df["logp"] <= 5)
+        ]
+        print(f"Lipinski filter: {before} → {len(df)} compounds")
+
+    # Step 3: Sort by similarity
+    df["similarity"] = pd.to_numeric(df["similarity"], errors="coerce")
+    df = df.sort_values("similarity", ascending=False)
+
+    print(f"VS prep: {len(df)} compounds ready for screening")
+    return df
+```
+
+## References
+
+### Output Files
+
+| ファイル | 形式 |
+|---|---|
+| `results/zinc_search.csv` | CSV |
+| `results/zinc_similar.csv` | CSV |
+| `results/zinc_substance.json` | JSON |
+| `results/zinc_catalogs.csv` | CSV |
+| `results/vs_library.csv` | CSV |
+
+### 利用可能ツール
+
+| カテゴリ | 主要ツール | 用途 |
+|---|---|---|
+| ZINC | `ZINC_search_by_name` | 化合物名検索 |
+| ZINC | `ZINC_search_by_smiles` | SMILES 類似性検索 |
+| ZINC | `ZINC_get_substance` | 化合物詳細 |
+| ZINC | `ZINC_get_catalogs` | カタログ一覧 |
+
+### 参照スキル
+
+| スキル | 関連 |
+|---|---|
+| `scientific-compound-similarity` | 化合物類似性 |
+| `scientific-pharmacology-targets` | 薬理学ターゲット |
+| `scientific-molecular-docking` | 分子ドッキング |
+| `scientific-drug-target-interaction` | DTI 解析 |
+| `scientific-admet-toxicity` | ADMET 毒性 |
+
+### 依存パッケージ
+
+`requests`, `pandas`