htlin222
meta-pipe
Python

Claude Code-powered end-to-end meta-analysis automation: AI-assisted literature review, screening, extraction, analysis, and manuscript generation for systematic reviews and clinical evidence synthesis

Last updated Jul 6, 2026
106
Stars
40
Forks
10
Issues
+7
Stars/day
Attention Score
46
Language breakdown
Python 49.7%
TeX 28.4%
R 17.6%
Shell 4.3%
โ–ธ Files click to expand
README

Meta-Analysis Pipeline

GitHub stars Last commit License Python

AI-assisted, end-to-end meta-analysis with reproducible tooling.

From research question to manuscript-ready output -- powered by Claude Code.


Quick Start

# 1. Setup (one-time)
cp .env.example .env        # Add your API keys
cd tooling/python && uv init

2. Create a new project

uv run tooling/python/init_project.py --name my-meta-analysis

3. Edit your research question

Open projects/my-meta-analysis/TOPIC.txt and paste your topic

4. Launch Claude Code and say:

"Start project my-meta-analysis" or "See projects/my-meta-analysis/TOPIC.txt and start"

That's it. Claude will handle the rest.

Don't have a topic yet? Say:

"Help me brainstorm a topic"

Claude will guide you through an interactive session to develop your research question and create the project for you.


What Claude Does

  • Creates your project in projects/<your-project-name>/
  • Reads your topic from projects/<your-project-name>/TOPIC.txt
  • Asks clarifying questions (databases, outcomes, dates)
  • Runs the 9-stage pipeline automatically
  • Generates manuscript-ready outputs

Pipeline Overview

All stages are created inside projects/<your-project-name>/:

| Stage | Output | | ------------- | ------------------------- | | 01_protocol | pico.yaml, eligibility.md | | 02_search | dedupe.bib | | 03_screening | decisions.csv | | 04_fulltext | manifest.csv | | 05_extraction | extraction.csv | | 06_analysis | figures/, tables/ | | 07_manuscript | manuscript.pdf | | 08reviews | gradesummary.md | | 09qa | finalqa_report.md |


Example: Completed Project

See a real meta-analysis โ†’ projects/ici-breast-cancer/

This is a 99% complete meta-analysis on immune checkpoint inhibitors in triple-negative breast cancer:

  • 5 RCTs, N=2,402 patients
  • Primary outcome: RR 1.26 (95% CI 1.16-1.37), p=0.0015, โŠ•โŠ•โŠ•โŠ• HIGH quality
  • Manuscript: 4,921 words (Lancet Oncology compliant)
  • Time invested: ~14 hours (vs 100+ hours manual)
Quick tour:
  • projects/ici-breast-cancer/README.md - Project overview
  • projects/ici-breast-cancer/00overview/FINALPROJECT_SUMMARY.md - Key findings
  • projects/ici-breast-cancer/07_manuscript/ - Complete manuscript (5 sections)
Use as template for your own meta-analysis workflow.

Project Structure

meta-pipe/
โ”œโ”€โ”€ ma-*/                    # Framework code modules
โ”œโ”€โ”€ docs/archive/            # Archived documentation
โ”œโ”€โ”€ tooling/                 # Shared tools and scripts
โ””โ”€โ”€ projects/                # All your meta-analysis projects
    โ”œโ”€โ”€ ici-breast-cancer/   # Example: complete meta-analysis
    โ”œโ”€โ”€ legacy/              # Historical data (pre-2026-02-08)
    โ””โ”€โ”€ your-project/        # Your new projects
        โ”œโ”€โ”€ 01_protocol/
        โ”œโ”€โ”€ 02_search/
        โ”œโ”€โ”€ ...
        โ”œโ”€โ”€ 09_qa/
        โ””โ”€โ”€ TOPIC.txt

Note: Projects are isolated and NOT tracked by Git (see .gitignore). Only the example project ici-breast-cancer/ is tracked for reference.


Documentation

| Doc | Purpose | | ----------------------------------------------------------- | ---------------------------- | | GETTING_STARTED.md | Manual step-by-step guide | | API Setup | Database API keys | | CLAUDE.md | Agent behavior (auto-loaded) | | tooling/scripts/ | Utility scripts |

Citation

If you use meta-pipe in your research, please cite it:

AMA Format:

Lin HT, Yeh JT. meta-pipe: AI-assisted, end-to-end meta-analysis pipeline with reproducible tooling. GitHub; 2025. Accessed 2026. https://github.com/htlin222/meta-pipe

BibTeX:

@software{linmetapipe2025,
  author       = {Lin, Hsieh-Ting and Yeh, Jiunn-Tyng},
  title        = {meta-pipe: AI-assisted, end-to-end meta-analysis pipeline with reproducible tooling},
  year         = {2025},
  url          = {https://github.com/htlin222/meta-pipe},
  note         = {Accessed: 2026}
}

Requirements

  • uv (Python)
  • R โ‰ฅ 4.2 + renv
  • cmake (required for building R packages like fs on macOS ARM)
  • Quarto
  • API keys in .env

Citation

If you use meta-pipe in your research, please cite it.

AMA format:

Lin HT, Yeh JT. meta-pipe: AI-assisted, end-to-end meta-analysis pipeline with reproducible tooling. GitHub; 2026. Accessed March 22, 2026. https://github.com/htlin222/meta-pipe

BibTeX:

@software{lin2026metapipe,
  author    = {Lin, Hsieh-Ting and Yeh, Jiunn-Tyng},
  title     = {meta-pipe: AI-Assisted, End-to-End Meta-Analysis Pipeline with Reproducible Tooling},
  year      = {2026},
  url       = {https://github.com/htlin222/meta-pipe},
  note      = {Software}
}

ยฉ 2026 GitRepoTrend ยท htlin222/meta-pipe ยท Updated daily from GitHub