Skip to content

CALL_FOR_DATA.md

Latest commit

 

History

History
209 lines (146 loc) · 8.96 KB

CALL_FOR_DATA.md

File metadata and controls

209 lines (146 loc) · 8.96 KB
Raw

📢 Call for Longitudinal Pathology & Omics Data

Benchmarking Dynamical Disease Frameworks — Project Confluence

Status: Accepting Submissions

Objective

Project Confluence is validating a universal dynamical systems framework that models disease progression and therapeutic response based on complexity profiles rather than disease-specific labels.

We seek wet-lab validation datasets to test whether shared mathematical structures underlie therapeutic success across distinct pathologies — Oncology, Metabolic Disease, and Comorbidities.

Core Hypothesis: Disease is a loss of regulatory complexity (a shift to a pathological attractor). Therapeutic success is a restoration of healthy dynamics. If true, the same mathematics should identify therapeutic windows in both a tumor regression dataset and a glycemic control dataset.

What We Mean by "Complexity"

Confluence does not analyze individual genes or proteins. It analyzes state-space trajectories — how a patient's measurable variables evolve over time. From these trajectories, we compute a 5-dimensional Φ (Phi) vector:

Φ Dimension What It Measures Computation Healthy Range
Φ_temporal Temporal regularity Multiscale Entropy (MSE) 0.6–0.8
Φ_spatial State-space dimensionality Correlation Dimension (D₂) 3.0–6.0
Φ_functional Perturbation recovery Recovery rate after stress 0.5–0.8
Φ_informational Predictability structure Lyapunov exponent + spectral slope 0.5–0.7
Φ_coupling Cross-system coherence Inter-variable correlation 0.4–0.7

A "therapeutic success" in Confluence terms is a measurable shift in Φ from a pathological archetype back toward the healthy complex attractor.

For full definitions, see validation/complexity_signature.md.

Data We Are Seeking

Required: Time-Series / Longitudinal Data

Important

Static snapshots (single biopsy, single blood draw) cannot reveal complexity profiles. We need multiple timepoints per patient, ideally spanning pre-intervention → intervention → post-intervention.

1. Separate Pathologies

Oncology Datasets

  • Longitudinal tumor measurements (imaging, liquid biopsy, ctDNA)
  • Multi-timepoint omics (RNA-seq, proteomics, metabolomics)
  • Treatment response outcomes (RECIST, pathological response, survival)
  • Any cancer type — we are disease-agnostic

Metabolic Disease Datasets

  • Longitudinal metabolic markers (HbA1c, fasting glucose, insulin, C-peptide)
  • Continuous glucose monitoring (CGM) traces (ideal for entropy calculation)
  • Intervention outcomes (medication changes, lifestyle modifications, bariatric surgery)
  • Type 1, Type 2, MODY, or gestational diabetes

2. Conjoined Pathologies (Comorbidity)

  • Patients with both malignancy and metabolic dysfunction
  • Data tracking how intervention in one domain affects the other
  • Examples: metformin use in breast cancer, diabetes onset post-chemotherapy, pancreatic cancer with glucose dysregulation

3. Negative Results

Tip

Data where therapeutic value was not found is equally valuable. Failed interventions define the boundaries of pathological attractor basins and help us characterize treatment-resistant complexity states.

Ideal Dataset Features

Feature Priority Why It Matters
≥3 timepoints per patient 🔴 Critical Minimum for trajectory reconstruction
Pre- and post-intervention 🔴 Critical Needed to measure Φ shift
Sampling frequency recorded 🟡 High Required for entropy calculation
Multi-modal (omics + clinical) 🟡 High Enables cross-domain Φ_coupling
Intervention metadata (dose, timing, type) 🟡 High Required for perturbation modeling
Raw measurements (not just summary stats) 🟢 Preferred Enables full complexity profiling
≥20 patients per cohort 🟢 Preferred Statistical power for archetype classification

Metadata Schema

For each submitted dataset, please provide these fields (see validation/data_submission_template.json for the full schema):

{
  "dataset_name": "...",
  "pathology_class": "oncology | metabolic | comorbidity",
  "sampling_frequency": "daily | weekly | monthly | irregular",
  "n_timepoints_per_patient": "median value",
  "intervention_timestamp": "relative day/hour of intervention start",
  "outcome_trajectory": "continuous | categorical",
  "variables_measured": ["list of measured quantities"],
  "variable_variance_available": true
}

What We Offer in Return

🔬 Complexity Analysis (Free)

We will run the full Confluence pipeline on your dataset and return:

  • Φ profile for each patient (5D complexity vector over time)
  • Archetype classification (Chaotic/Decoupled, Rigid/Locked, Collapsed/Exhausted)
  • Complexity recovery score (pre→post intervention Φ shift)
  • Attractor landscape visualization (phase portraits, bifurcation diagrams)

📝 Validation Partnership

  • Co-authorship on publications demonstrating cross-disease dynamical universality
  • Named acknowledgment in the Confluence validation registry
  • Priority access to framework updates and analytical tools

📦 FAIR Data Support

  • Assistance depositing data in compliant repositories (GEO, NCI IDC, NIDDK Central Repository)
  • DOI generation for your contribution
  • Metadata standardization to LOINC/SNOMED-CT codes

How to Submit

Option A: Full Dataset Sharing

Share your dataset directly (CSV, HDF5, or database export) with metadata using our submission template.

📧 Contact: [kelechi@projectconfluence.org]
🔗 Upload: GitHub Issues

Option B: Federated Complexity Analysis

If data cannot leave your institution, we provide the Confluence Complexity Script — a self-contained Python package. You run it locally and send us only the anonymized Φ profiles (no patient-level data leaves your system).

# Install
pip install confluence-profiler  # (coming soon)

# Run
python -m confluence.federated --input your_data.csv --output phi_profiles.json

Option C: Existing Public Data Nomination

Know of a public dataset (GEO, TCGA, NIDDK, All of Us) that meets our criteria? Open an issue or email us the accession number — we'll handle the rest.

Target Data Sources

Repository Disease Area URL
NCI Imaging Data Commons (IDC) Cancer pathology + imaging portal.imaging.datacommons.cancer.gov
NIDDK Central Repository Diabetes & metabolic repository.niddk.nih.gov
All of Us Research Program Comorbidity (cancer + metabolic) researchallofus.org
GEO / ArrayExpress Multi-omics time-series ncbi.nlm.nih.gov/geo
TCGA (via GDC) Pan-cancer omics portal.gdc.cancer.gov
PhysioNet Physiological time-series physionet.org

Validation Arms

This data call supports a three-arm validation strategy (full protocol: validation/validation_protocol.md):

Arm Goal Success Metric
Separate Confluence works on Cancer and Diabetes individually Same equations identify therapeutic tipping points in both
Conjoined Confluence handles coupled disease systems Model predicts cross-domain interaction effects
Universality The mathematics is disease-agnostic Complexity recovery profiles are statistically indistinguishable across diseases

Timeline

Milestone Target Date
Call for Data posted March 2026
First public dataset ingested April 2026
Separate Arm validation complete June 2026
Conjoined Arm pilot August 2026
Universality analysis preprint Q4 2026

Framework Preview

Citation

If you use Project Confluence or reference this Call for Data:

@software{ogbonna2026confluence,
  author = {Ogbonna, Kelechi},
  title = {Project Confluence: Universal Dynamical Framework for Therapeutic Validation},
  year = {2026},
  url = {https://github.com/cloudynirvana/project-confluence}
}

"The measure of health is not the absence of disease, but the presence of complexity."