PRISM: privacy-preserving rare disease analysis using fully homomorphic encryption.

Area of Science:

• Genomics
• Bioinformatics
• Privacy-Preserving Technologies

Background:

• Rare diseases impact millions globally, but genomic analysis is hindered by limited data and strict privacy regulations like GDPR.
• Collaborative genomic data analysis is crucial for understanding rare disease mechanisms but faces significant privacy challenges.

Purpose of the Study:

• To introduce PRISM, a novel privacy-preserving framework for rare disease variant analysis.
• To enable secure, cross-institutional collaboration on sensitive genomic data.
• To facilitate variant filtering across recessive, dominant, and de novo inheritance models without data exposure.

Main Methods:

• Development of PRISM, a framework utilizing threshold fully homomorphic encryption (FHE) for decentralized key management.
• Implementation of two algorithmic variants: multiplication-intensive (MUL-IN) and addition-intensive (ADD-IN).
• Filtering of disease-causing variants on encrypted data using FHE.

Main Results:

• PRISM enables rare disease variant analysis across multiple institutions while preserving data privacy.
• The ADD-IN algorithms achieve significant runtime improvements (up to 17x for recessive/dominant, 22x for de novo) compared to MUL-IN.
• The framework demonstrates practicality and scalability for analyzing millions of variants in a single-cloud setting.

Conclusions:

• PRISM offers a groundbreaking solution for privacy-preserving rare disease genomic analysis.
• The framework overcomes data access limitations imposed by privacy regulations.
• This study pioneers the use of FHE for multi-inheritance model rare disease variant analysis.