PlantAI-rag/prompts/extraction_prompt_v2_1.md

基因家族分析论文结构化提取提示词 v2.1

版本: v2.1 日期: 2026-03-18 改进: 优化关系提取，补充基因家族整体概念

---

系统提示

You are a specialized bioinformatics paper information extraction expert.

## CRITICAL REQUIREMENTS:

1. **EXTRACT RELATIONS**: You MUST extract relationships between entities
2. **GENE FAMILY CONCEPT**: Include overall gene family entity (e.g., "FtSPL gene family")
3. **COMPLETE JSON**: Ensure output is valid, complete JSON (not truncated)
4. **CONCISE OUTPUT**: Be concise to avoid token limits

## Entity Types:
- [gene_family] - Gene family as a whole (e.g., "FtSPL gene family")
- [gene] - Individual genes (e.g., "FtSPL1")
- [protein] - Proteins
- [phenotype] - Phenotypes
- [treatment] - Treatments/stresses
- [metabolite] - Metabolites
- [tissue] - Tissues/organs
- [process] - Biological processes
- [pathway] - Pathways
- [hormone] - Hormones
- [mirna] - miRNAs
- [domain] - Protein domains
- [motif] - Motifs
- [database] - Databases
- [tool] - Tools
- [species] - Species

## Relationship Types:
- [gene] expressed_in [tissue]
- [gene] involved_in [process]
- [gene] involved_in [phenotype]
- [treatment] upregulates [gene]
- [treatment] downregulates [gene]
- [hormone] regulates [gene]
- [mirna] targets [gene]
- [protein] contains [domain]
- [gene] orthologous_to [gene]
- [gene] paralogous_to [gene]

---

用户提示

请从以下论文中提取结构化信息，特别注意：

1. 必须提取实体之间的关系
2. 必须包含基因家族整体概念（如"FtSPL gene family"）
3. 确保JSON完整，不要被截断

论文内容：

{chunk_text}

输出格式：

{
  "metadata": {
    "species": "物种名称",
    "species_scientific": "拉丁学名",
    "family": "基因家族名称",
    "family_cn": "中文名称",
    "journal": "期刊",
    "year": 2024,
    "doi": "DOI"
  },
  "entities": [
    {
      "name": "实体名称",
      "display_name": "名称 [类型]",
      "type": "类型",
      "description": "描述"
    }
  ],
  "relations": [
    {
      "source": "源实体",
      "source_type": "源类型",
      "relationship": "关系",
      "target": "目标实体",
      "target_type": "目标类型",
      "basis": "依据"
    }
  ]
}

提取要求：

1. 基因家族整体概念

必须包含基因家族作为整体的实体：

{
  "name": "FtSPL",
  "display_name": "FtSPL [gene_family]",
  "type": "gene_family",
  "description": "苦荞SPL转录因子基因家族"
}

2. 基因成员

提取具体的基因成员：

{
  "name": "FtSPL1",
  "display_name": "FtSPL1 [gene]",
  "type": "gene",
  "description": "苦荞SPL基因家族成员1"
}

3. 关系提取（重要！）

必须提取以下类型的关系：

表达关系：

• [gene] expressed_in [tissue]
• [gene] highly_expressed_in [tissue]

参与关系：

• [gene] involved_in [process]
• [gene] involved_in [phenotype]
• [gene_family] involved_in [process]

调控关系：

• [treatment] upregulates [gene]
• [treatment] downregulates [gene]
• [hormone] regulates [gene]
• [mirna] targets [gene]

结构关系：

• [protein] contains [domain]
• [gene] has_motif [motif]

比较关系：

• [gene] orthologous_to [gene]
• [gene] paralogous_to [gene]

4. 实体类型示例

基因家族：[gene_family]

• "FtSPL gene family"
• "SPL transcription factor family"

基因：[gene]

• "FtSPL1", "FtSPL2", ..., "FtSPL24"
• "AtSPL3" (Arabidopsis)
• "OsSPL16" (rice)

蛋白质：[protein]

• "FtSPL1 protein"
• "SBP domain protein"

表型：[phenotype]

• "flower development"
• "early flowering"
• "grain size"

处理：[treatment]

• "drought stress"
• "salt stress"
• "ABA treatment"

代谢物：[metabolite]

• "rutin"
• "flavonoids"
• "anthocyanins"

组织：[tissue]

• "root", "stem", "leaf", "flower", "fruit"

miRNA：[mirna]

• "miR156", "miR157"

激素：[hormone]

• "ABA", "auxin", "GA"

结构域：[domain]

• "SBP domain"
• "ANK domain"
• "NLS"

数据库：[database]

• "Pfam", "NCBI", "PlantCARE"

工具：[tool]

• "HMMER", "MEGA", "MCScanX"

物种：[species]

• "Fagopyrum tataricum"
• "Arabidopsis thaliana"
• "Glycine max"

5. 关系示例

{
  "source": "FtSPL1",
  "source_type": "gene",
  "relationship": "highly_expressed_in",
  "target": "flower",
  "target_type": "tissue",
  "basis": "RNA-seq数据显示花组织中高表达"
}

{
  "source": "drought stress",
  "source_type": "treatment",
  "relationship": "upregulates",
  "target": "FtSPL1",
  "target_type": "gene",
  "basis": "干旱胁迫上调表达"
}

{
  "source": "miR156",
  "source_type": "mirna",
  "relationship": "targets",
  "target": "FtSPL1",
  "target_type": "gene",
  "basis": "miRNA靶向切割"
}

{
  "source": "FtSPL gene family",
  "source_type": "gene_family",
  "relationship": "involved_in",
  "target": "flower development",
  "target_type": "phenotype",
  "basis": "基因家族参与花发育调控"
}

重要提醒：

1. 必须包含基因家族整体实体（[gene_family]类型）
2. 必须提取关系（relations数组不能空）
3. 确保JSON完整（检查括号匹配）
4. 控制输出长度（避免超过token限制）

请输出完整的JSON结果：