BioAnchor 🔬⚓

The Reproducibility & Provenance Layer for Bayesian Scientific Analysis

BioAnchor is an autonomous scientific agent that executes Bayesian MCMC analyses, quantifies uncertainty, and permanently archives verifiable evidence on decentralized storage — creating a cryptographic chain-of-trust for scientific discovery.

---

Why BioAnchor?

Science has a reproducibility crisis. Over 70% of researchers have failed to reproduce another scientist's experiments (Nature, 2016). The root causes: lost data, opaque methods, and no verifiable provenance chain.

BioAnchor solves this by combining three capabilities that no other tool offers together:

• Autonomous Bayesian Analysis — Runs PyMC 5 / NUTS MCMC models with full posterior sampling and convergence diagnostics
• Uncertainty Quantification — Computes coefficient of variation (CV), R-hat, ESS, and credible intervals for every parameter
• Permanent Decentralized Archiving — Stores complete MCMC metadata (posteriors, diagnostics, model specs) on Arweave for immutable, censorship-resistant provenance

How It Works

Raw Data → BioAnchor Agent → MCMC Analysis → Uncertainty Scores → Arweave Archive
                                                                      ↓
                                                              Permanent TX ID
                                                         (verifiable by anyone)

1. Ingest — Feed dose-response, genomic, or any biological dataset
2. Analyze — BioAnchor autonomously configures and runs Bayesian MCMC (PyMC 5, NUTS sampler)
3. Quantify — Generates uncertainty metrics (CV, posterior distributions, convergence diagnostics)
4. Archive — Permanently stores analysis metadata on Arweave with a unique Transaction ID
5. Verify — Anyone can retrieve and reproduce the analysis using the TX ID

Quick Start

pip install bioanchor

from bioanchor import BioAnchor

# Initialize with Arweave wallet
anchor = BioAnchor(wallet_path="arweave-wallet.json")

# Run MCMC analysis and archive results
result = anchor.analyze_and_archive(
    data=my_dataset,
    model_type="hill_equation",  # IC50 dose-response model
    n_samples=2000,
    n_chains=4
)

# Get permanent Arweave TX ID
print(f"Archived: {result.tx_id}")
print(f"Verify at: https://ar-io.dev/{result.tx_id}")

Real-World Validation

BioAnchor has been validated with real pharmaceutical data:

• IC50 dose-response modeling using Hill equation with PyMC 5
• 4-chain NUTS sampling with full convergence diagnostics
• Live Arweave archive: 2lIZtjsu120ERqPKu6XbITpUnAveWeH7hI4owBfNCdY

Architecture

┌─────────────────────────────────────────────────┐
│                 BioAnchor Agent                  │
├─────────────────────────────────────────────────┤
│  ┌───────────┐  ┌──────────┐  ┌──────────────┐ │
│  │  PyMC 5   │  │ ArviZ    │  │  Uncertainty │ │
│  │  Engine   │→ │Diagnostics│→│  Scoring     │ │
│  │ (NUTS)    │  │ (R-hat,  │  │  (CV, CI,    │ │
│  │           │  │  ESS)    │  │   posterior)  │ │
│  └───────────┘  └──────────┘  └──────┬───────┘ │
│                                       │         │
│  ┌────────────────────────────────────▼───────┐ │
│  │         Decentralized Storage Layer        │ │
│  │  ┌──────────┐  ┌──────────┐  ┌──────────┐ │ │
│  │  │ Arweave  │  │ Filecoin │  │  IPFS    │ │ │
│  │  │ (live)   │  │ (planned)│  │(planned) │ │ │
│  │  └──────────┘  └──────────┘  └──────────┘ │ │
│  └────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────┘

DeSci Integration Roadmap

BioAnchor is designed to be the scientific truth verification layer for the decentralized science (DeSci) ecosystem:

Phase 1: Multi-Chain Storage (Current → Q3 2026)

• ✅ Arweave permanent archiving (live)
• 🔄 Filecoin/IPFS bridge (in development)
• 🔄 Multi-chain verification endpoints

Phase 2: IP-NFT & Data Quality Oracle (Q3–Q4 2026)

• Uncertainty-weighted reliability scoring for IP-NFTs
• Integration with Molecule / BIO Protocol IP framework
• CV-based data quality oracle for BioDAOs (VitaDAO, AthenaDAO, etc.)

Phase 3: Autonomous Bayesian Scientist Agent (Q4 2026+)

• AI-driven hypothesis generation from literature
• Automated MCMC model selection and execution
• Self-archiving results with provenance chain
• Integration with BIO Protocol Scientific AI Agent platform

Phase 4: DAO Governance & Token Economy (2027)

• Community-governed research prioritization
• Staking on reproducibility verification
• Token-incentivized peer review of MCMC analyses

For BioDAOs & DeSci Projects

BioAnchor provides a trust layer that any BioDAO can plug into:

Use Case	How BioAnchor Helps
Drug Discovery	Verify IC50/EC50 dose-response analyses with full uncertainty quantification
IP-NFT Valuation	Score data quality using CV metrics before tokenizing research IP
Reproducibility Audits	Retrieve and re-run any archived MCMC analysis from its Arweave TX ID
Grant Accountability	Prove research outputs are real and verifiable on-chain

Testing

BioAnchor includes a MockUploader for development and testing without Arweave wallet:

from bioanchor import BioAnchor, MockUploader

anchor = BioAnchor(uploader=MockUploader())
result = anchor.analyze_and_archive(data=test_data)
# Returns simulated TX ID for testing

Tech Stack

• Python 3.11+ with PyMC 5, ArviZ, SciPy
• Arweave via ar-io.dev gateway (Node.js uploader for production)
• Zenodo DOI: 10.5281/zenodo.19709077

Citation

If you use BioAnchor in your research, please cite:

@software{kim2026bioanchor,
  author       = {Kim, Dohoon},
  title        = {BioAnchor: Bridging Bayesian Biological Data Analysis with Decentralized Storage},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.19709077},
  url          = {https://doi.org/10.5281/zenodo.19709077}
}

About

BioAnchor is developed by Dohoon Kim at Promptgenix LLC, building open-source scientific infrastructure for verifiable, reproducible research.

• 📦 PyPI: pip install bioanchor
• 🔗 GitHub: github.com/kdh4win4/bioanchor
• 📄 Paper: Submitted to Bioinformatics (Oxford) — BIOINF-2026-1271

License

MIT License — see LICENSE for details.