BiomarkerExtract

AI-Powered Biomarker Discovery from Scientific Literature

Automated extraction and validation of aging biomarkers using state-of-the-art Large Language Models

Features • Quick Start • Documentation • Results • Citation

---

Contact

Feel free to contact me via email for any needs: [email protected]

---

Overview

BiomarkerExtract is a production-ready pipeline for discovering and validating aging biomarkers from scientific literature using Large Language Models. Built on Google's LangExtract framework, it supports multiple LLM providers and delivers publication-quality results at ultra-low cost.

Key Achievements

• ✅ 79 biomarkers extracted from 46 scientific papers
• ✅ 93.7% validation rate with scientific evidence
• ✅ 84.8% high confidence (≥0.90) extractions
• ✅ $0.003 per paper processing cost
• ✅ 5 LLM providers supported out-of-the-box

---

Features

Multi-Provider LLM Support

• OpenRouter - 100+ models with single API key (Recommended)
• OpenAI - GPT-5.2, GPT-4o, O1
• Anthropic - Claude 4.5, Sonnet 4.5
• Google - Gemini 3.0, Gemini Pro
• Ollama - Local inference (FREE)

Complete Pipeline

1. Literature Search - PubMed + bioRxiv integration
2. Biomarker Extraction - LLM-powered entity recognition
3. Scientific Validation - Automated quality assessment
4. Multi-Format Export - JSON, CSV, TXT

Analysis & Visualization

• Publication-quality charts
• Network analysis
• Category distribution
• Confidence metrics
• Sample size statistics

---

Quick Start

Installation

# Clone repository
git clone https://github.com/AntonioVFranco/BiomarkerExtract.git
cd BiomarkerExtract

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run installation
bash install_unified.sh

Configuration

# Set your API key (choose one provider)
export OPENROUTER_API_KEY="sk-or-v1-xxxxxxxx"  # Recommended
# OR
export OPENAI_API_KEY="sk-xxxxxxxx"
# OR
export ANTHROPIC_API_KEY="sk-ant-xxxxxxxx"

Run Pipeline

# Quick start with OpenRouter (cheapest)
bash run_openrouter.sh

# Or with other providers
bash run_openai.sh      # OpenAI GPT-5.2
bash run_anthropic.sh   # Claude 4.5
bash run_ollama.sh      # Local (FREE)

Python API

from langextract.providers import unified_production_pipeline as upp

results = upp.run_pipeline(
    biomarker_terms=["Horvath clock", "GDF-15", "NAD+"],
    pubmed_email="[email protected]",
    provider="openrouter",
    api_key="your-key",
    max_papers=20
)

print(f"Extracted {results['statistics']['biomarkers_extracted']} biomarkers!")

---

Results

Sample Extraction (v0.1)

Metric	Value
Papers Processed	46
Biomarkers Extracted	79
Validated	74 (93.7%)
High Confidence	67 (84.8%)
Processing Time	8.35 minutes
Total Cost	$0.15

Top Biomarkers Discovered

1. Horvath clock (12 mentions) - Epigenetic
2. GDF-15 (9 mentions) - Proteomic
3. NAD+ levels (3 mentions) - Metabolomic
4. Hannum clock (2 mentions) - Epigenetic
5. DunedinPACE (2 mentions) - Epigenetic

Category Distribution

• Epigenetic: 43.0%
• Proteomic: 34.2%
• Cellular: 10.1%
• Metabolomic: 6.3%
• Genomic: 2.5%
• Transcriptomic: 2.5%

---

Documentation

• UNIFIED_README.md - Complete system overview
• UNIFIED_CONFIGURATION.md - Provider setup guides
• Phase3_README.md - Core biomarker models
• Phase4_README.md - Literature pipeline
• Option2_Testing_README.md - Testing suite

Examples

See examples_unified.py for 9 complete working examples:

• Basic extraction
• Batch processing
• Provider comparison
• Custom models
• Complete pipeline

---

Cost Comparison

Processing 1000 papers:

Provider	Cost	Speed	Accuracy
OpenRouter	$3.00 ⭐	Fast	88%
OpenAI GPT-5.2	$25.00	Fast	92%
Anthropic Claude 4.5	$30.00	Medium	90%
Google Gemini 3.0	$10.00	Fast	85%
Ollama	FREE ⭐	Slow*	80%

*Depends on local GPU

---

Architecture

BiomarkerExtract/
├── langextract/
│   ├── core/
│   │   └── biomarker_models.py      # 21 Pydantic models
│   ├── literature/
│   │   ├── pubmed_client.py         # PubMed API
│   │   ├── biorxiv_client.py        # bioRxiv API
│   │   ├── pdf_parser.py            # PDF extraction
│   │   └── batch_processor.py       # Parallel processing
│   └── providers/
│       ├── unified_llm_provider.py           # 5 LLM providers
│       └── unified_production_pipeline.py    # End-to-end pipeline
├── tests/
│   └── option2/                     # 30+ tests (93% passing)
├── examples_unified.py              # 9 working examples
└── run_*.sh                         # Quick-start scripts

---

Testing

# Run complete test suite
cd tests/option2
bash full_test.sh

# Quick validation
bash quick_test.sh

# Results: 93% tests passing

---

Statistics

• ~5,700 lines of production code
• 21 Pydantic models for data validation
• 5 LLM providers integrated
• 30+ unit tests (93% success rate)
• 3 formats for data export (JSON, CSV, TXT)
• Publication-quality visualizations included

---

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

---

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Based on Google's LangExtract framework.

---

Citation

If you use BiomarkerExtract in your research, please cite:

@software{biomarkerextract2026,
  author = {Franco, Antonio V.},
  title = {BiomarkerExtract: AI-Powered Biomarker Discovery from Scientific Literature},
  year = {2026},
  version = {0.1},
  url = {https://github.com/AntonioVFranco/BiomarkerExtract}
}

---

Acknowledgments

• Built on Google's LangExtract
• Inspired by aging research and longevity science
• Powered by state-of-the-art Large Language Models

---

Connect

• Email: [email protected]
• Medium: @AntonioVFranco
• LinkedIn: antoniovfranco
• GitHub: @AntonioVFranco

---

⭐ Star this repo if you find it useful! ⭐

Report Bug • Request Feature • Documentation