sbaynes presents research from Microsoft, Drexel, and the Broad Institute on leveraging generative AI—such as Azure AI Foundry—to support genetic professionals in diagnosing rare diseases and addressing challenges in whole genome sequencing analysis.

Using AI to Assist in Rare Disease Diagnosis: Insights from Microsoft Research

In the rapidly advancing field of genetic analysis, diagnosing rare diseases requires the synthesis of vast, complex data from whole genome sequencing. However, fewer than half of initial cases yield a diagnosis, and reanalyzing cases as new scientific findings emerge remains complex and time-consuming for genetic professionals.

Microsoft Research’s Comprehensive Study

A team from Microsoft Research, in collaboration with Drexel University and the Broad Institute, conducted an in-depth study—AI-Enhanced Sensemaking: Exploring the Design of a Generative AI-Based Assistant to Support Genetic Professionals—published in ACM Transactions on Interactive Intelligent Systems. The study investigates how generative AI can help overcome core challenges faced in rare disease diagnosis and genetic data interpretation.

Key Challenges Identified

  • Information Overload: Analysts must filter and synthesize data from numerous sources, which is laborious and prone to error.
  • Collaborative Sharing: Sharing and jointly interpreting findings is often inefficient due to outdated processes.
  • Prioritizing Reanalysis: The influx of new scientific evidence makes it difficult to systematically determine which unsolved cases merit reanalysis.

Co-Designing an AI Assistant

The study involved interviews and co-design workshops with 17 genetics professionals to identify workflows where AI assistance is needed. Two high-priority tasks emerged:

  1. Flagging Cases for Reanalysis: The AI assistant should proactively alert professionals to unsolved cases that might benefit from new discoveries.
  2. Aggregating and Synthesizing Evidence: The tool should automatically gather and summarize gene/variant data from scientific literature to save analysts time and support thorough reviews.

Participants also prioritized collaborative features for interpreting, editing, and verifying AI-generated artifacts.

Integrating Azure AI Foundry and Microsoft Technologies

The Azure AI Foundry Labs showcase how experimental AI technologies can be integrated into real-world workflows. The prototype assistant demonstrated at these labs aimed to:

  • Provide flexible filtering and deep evidence aggregation
  • Enable collaborative editing and verification of AI-generated data
  • Synthesize information from multiple modalities (e.g., text, imagery, structured data)

Design Implications for Expert-AI Sensemaking

Three main design considerations informed the AI assistant:

  • Distributed Sensemaking: Ensure artifacts can be collectively reviewed and annotated, fostering trust and reducing errors.
  • Dynamic, Temporal Support: Artifacts must capture prior rationale and keep track of new insights as scientific knowledge evolves.
  • Multimodal Evidence Integration: Support the combination of diverse data types to produce comprehensive, actionable reports for diagnosis.

Outlook and Next Steps

While the research prototype shows promise—increasing diagnostic yield and reducing the time to diagnosis—the team emphasizes the need for further, real-world evaluation. Success requires collaboration among AI developers, domain experts, and HCI researchers to ensure trustworthy, usable, and targeted solutions.

For those interested in further reading and engaging with this work, see:

The article encourages community engagement and continued exploration of AI’s transformative potential in life sciences.

This post appeared first on “Microsoft News”. Read the entire article here