Gradient Biotech
U

Menu

Research Areas

Computational BiologySingle-cell, bulk RNA-seq, spatial, and biomarkers
CardiologyCardiovascular signals, cohorts, and interpretationImmunologyImmune cell states, signaling networks, and repertoire dynamicsNeurologyBrain connectivity, functional imaging, and electrophysiologyOncologyTumor microenvironment, immune response, mutations, and survivalPathologyWhole-slide imaging, tissue quantification, and spatial pathology

Application

Documentation

Documentation

Back to studies

FAQ

Is this a clinical diagnostic tool?

No. Cardiology in Gradient Biotech is a research and exploratory analysis workspace. Outputs must not be used for clinical decision-making. Interpretations include explicit disclaimers.

Do I need MATLAB or WFDB installed locally?

No.

Preprocessing failed — what should I check?

  • WFDB uploads include both .hea and .dat
  • CSV time column is numeric and strictly increasing
  • Recording is long enough (minimum row count enforced)
  • EDF channel index is valid for the file

Open the failed run's Logs tab for the specific error.

HRV button is disabled — why?

For ECG datasets, complete Preprocess ECG first. For RR CSV uploads, HRV should be available immediately after upload.

Interpret returns an error

Confirm Ollama credentials in backend-analysis/.env and that a completed metrics run exists as input. See repo root ai.md.

How do I compare treatment vs control?

  1. Import subjects with a treatment_group column
  2. Link datasets to subjects
  3. Run HRV (batch from cohort page if needed)
  4. Open Cohort → group by treatment_group → compare

Can I upload WFDB and CSV in the same batch?

No — one format per upload action. Create separate datasets for each recording.

Where are files stored locally?

Uploads: data/uploads/
Artifacts: data/artifacts/{run_id}/
Metadata: Supabase PostgreSQL (biochem_cardio_* tables) via the metadata API

These paths are gitignored.

How do I add a new pipeline?

Implement in backend-analysis/app/pipelines/, register in app/pipeline_registry.py, and expose via the existing job endpoints. See products/cardiology/plan.md and repo root claude.md for conventions.

What's the difference between descriptive and mechanistic interpret?

Descriptive summarizes metrics from one run or cohort comparison. Mechanistic proposes cross-domain physiological hypotheses when a subject has metrics from multiple modalities (for example HRV + BP + baroreflex).

Is cloud deployment supported?

Production deployment is supported on Google Cloud Run with Supabase for metadata. See deploy.sh and repo root README.md.