Exact Sciences (PreventionGenetics) accelerates variant curation and clinical note abstraction with Amazon Bedrock
Exact Sciences, through its PreventionGenetics subsidiary, uses AWS to accelerate variant curation and phenotype abstraction for genetic testing. The company manually reviewed scientific literature and patient clinical notes to interpret genetic variants that may cause rare disease or indicate elevated risk, and needed to speed turnaround for clinicians and patients. Working with the AWS Generative AI Innovation Center, Exact Sciences built the Variant Curation Accelerator and a phenotype abstraction tool on Amazon Bedrock, with human review, citations, source PDFs, and highlighted supporting text to improve trust and accuracy.
- Organization
- Exact Sciences
- Industry
- Healthcare
- Location
- United States
- Published
- May 2026
Reported outcomes
−30%
timeTime & speed
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1Document Intelligence
- 2Knowledge Extraction
- 3Clinical Decision Support
- Built the Variant Curation Accelerator and phenotype abstraction tool with the AWS Generative AI Innovation Center.
- Used Amazon Bedrock and Claude to search for relevant papers, analyze content, and present results in a UI for human review.
- Added guardrails including phenotype-match checks, citations, and source PDF highlighting for traceability.
- Used a machine learning document extraction capability for clinical notes to create markdown text for LLM input.
- At least 30% reduction in research times.
- Increased case review capacity by freeing genetic experts to focus on complex cases.
- Proof of concept completed in 6 weeks and tools spent about 1 year reaching production readiness.
Architecture
The solution uses Amazon Bedrock to run LLM analysis on searched scientific papers and clinical-note content, integrates outputs into a user interface via API for human review, and adds guardrails such as phenotype-matching checks, citations, and source-PDF highlighting. Clinical notes were converted into markdown text via an AWS document extraction capability before LLM processing.
Implementation partners2
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?