For Institutional Investors & Quant Teams

Epidemic Intelligence
as Alternative Data

Weekly ML outbreak probability scores across 200+ diseases and 9 geographies, mapped to 84 publicly traded tickers. Tamper-proof, backtested to 2010, machine-readable via API.

Request Data Access
84
disease-to-ticker mappings
176,800
disease-week observations in training data
2010
WHO DON history ingested from
89.0%
precision at β‰₯80% confidence threshold

What You Get

Three data products, one pipeline.

Signal Files

Weekly ML Probability Scores

ML outbreak probability for 200+ diseases across 9 geographies, updated weekly. Each file is SHA-256 hashed and archived to Cloudflare R2. Cryptographically verifiable, immutable record of every signal we've ever published.

  • β†’ 200+ diseases, 9 geographies
  • β†’ SHA-256 hash per file
  • β†’ R2-archived, auditable history
Ticker Mapping

84 Disease-to-Equity Links

Each disease is mapped to the pharma, biotech, and vaccine producer tickers most likely to reprice on an outbreak event. Covers vaccine developers, antiviral manufacturers, diagnostics companies, and PPE producers.

  • β†’ 84 tickers across disease categories
  • β†’ Pharma, biotech, diagnostics, PPE
  • β†’ Updated as coverage expands
API Access

Machine-Readable Feed

Pull scores directly into your quant models or data pipelines via the /v2/intelligence/ml-scores-with-tickers endpoint. Returns disease, geography, week, probability, confidence tier, and mapped tickers in a single response.

  • β†’ JSON response, versioned API
  • β†’ Probability + confidence tier
  • β†’ Tickers included per record

Signal Precision by Confidence Tier

Validated on held-out time periods. The model never saw this data during training.

Confidence Threshold
Precision
N Predictions
β‰₯50%
57.9%
106,959
β‰₯60%
62.9%
64,091
β‰₯70%
73.8%
23,199
β‰₯80% recommended
89.0%
7,185
β‰₯90% high-confidence
97.1%
2,192
Training data 176,800 disease-week observations. WHO DON back to 2010, FAO animal events back to 2020, CDC NNDSS, wastewater surveillance. 4+ years of validated signal history.
Validation method Time-series cross-validation. Model never sees future data during training. High-severity misses (Ebola, H5N1, plague) penalized 5Γ— more than false alarms.
Documented Case Study

Mpox 2022: 68-Day Lead on WHO PHEIC

Specific dates, documented signal, falsifiable. Not illustrative.

May 15, 2022

VW signal fires at 93.6%

Smallpox/Mpox probability crosses high-confidence threshold

May 23, 2022

First confirmed US case

8 days after VW signal

Jul 23, 2022

WHO declares PHEIC

68 days after VW signal. BVNRY surged ~3Γ— as market priced vaccine demand.

BVNRY share price vs VW Mpox signal, 68-day lead on WHO PHEIC

For institutional use only. Not investment advice.

Data Sources & Historical Depth

Our model trains on a multi-source dataset with years of validated history, not a few months of data.

β†’

WHO Disease Outbreak News

Ingested back to 2010. 15+ years of confirmed outbreak history across all geographies.

β†’

FAO Animal Disease Events

Back to 2020. Zoonotic spillover signals before human transmission is confirmed.

β†’

CDC NNDSS

US nationally notifiable disease surveillance data, updated continuously.

β†’

Wastewater Surveillance

Community-level signal 2-3 weeks before clinical case surges. Norovirus, COVID-19, Influenza, RSV, and more.

Request Institutional Data Access

We work directly with quant teams and institutional investors. Reach out with your use case and we'll walk through the data products, API, and backtest methodology.

Get in Touch

For institutional use only. Not investment advice. Signal files are for research and informational purposes.

Virus Watcher