Related Experiment Video
Updated: May 29, 2026
05:47
Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
The Evidence Aggregator: AI reasoning applied to rare disease diagnostics.
Hope Twede1, Lynn Pais2, Samantha Bryen3
1Microsoft Research, Microsoft Corporation, Redmond, WA, USA.
Summary
The Evidence Aggregator (EvAgg) tool uses generative AI to streamline rare disease diagnosis by quickly summarizing genetic variant information from scientific literature, reducing review time and improving diagnostic efficiency.
Area of Science:
- Genomics
- Medical Informatics
- Artificial Intelligence
Background:
- Rare disease diagnosis is complex and time-consuming, requiring extensive literature review.
- Synthesizing clinical and technical information for variant assessment is a significant bottleneck.
Purpose of the Study:
- To develop and evaluate the Evidence Aggregator (EvAgg), an AI-powered tool to aid rare disease diagnosis.
- To systematically extract and synthesize genetic variation data from scientific literature.
Main Methods:
- Developed EvAgg, an open-source, generative AI tool for extracting genetic variation information from literature.
- Created an expert-curated dataset for evaluating EvAgg's performance.
- Assessed EvAgg's accuracy in paper selection, variation detection, and data extraction.
- Conducted a user study to evaluate EvAgg's utility and user experience.
Main Results:
- EvAgg achieved 92% recall for relevant papers and 96% for genetic variation instances.
- Demonstrated approximately 80% accuracy in extracting case and variant-level content.
- Reduced literature review time by 34% and increased evaluation throughput.
Conclusions:
- EvAgg efficiently summarizes genetic variants and clinical features, supporting literature review for gene-disease relationships.
- The tool has the potential to decrease diagnostic delays and improve diagnostic success rates for rare diseases.
