AI

Agentic discovery of blood biomarker from distilled private health records

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/10/2026, 21:18:24 UTC.

Briefings

A privacy-preserving scoring tool was developed using Clalit Health Services data to evaluate blood biomarkers for 13 immune-mediated diseases. A graph attention network trained on private data predicted case-control AUCs, and the resulting weights were released. In external validation, agent-discovered biomarkers outperformed literature-seeded starting points by a median of 4.18 AUC percentage points across three independent cohorts.

Key points

• A graph attention network was trained on private Clalit Health Services data to predict case-control AUCs for 13 immune-mediated diseases.
• The released scoring tool improved external validation results by a median of 4.18 AUC percentage points compared to literature-seeded starting points.

Context

This study addresses the challenge of sharing private health records for frontier language model agents to discover new biomarkers, proposing a privacy-preserving approach that utilizes internal data validation.

Limitations

The tool relies on a single dataset from Clalit Health Services, and external validation was conducted across only three independent cohorts, limiting generalizability.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 24/10/2026, 01:11:28 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeicr5icimsakukcokmslctazdd5jtl7y4zehjdixahnnluyh4zjldm