Learning-augmented sketching offers improved performance for privacy preserving and secure GWAS.

Area of Science:

• Computational Biology
• Genomics
• Privacy-Preserving Technologies

Background:

• Trusted Execution Environments (TEEs) offer secure computation but face resource limitations.
• Genome-Wide Association Studies (GWAS) are crucial for genetic research but require significant computational resources.
• Existing methods like SkSES use sketching for privacy-preserving GWAS in TEEs, but accuracy can be limited.

Purpose of the Study:

• To develop a learning-augmented SkSES method for enhanced accuracy in GWAS within TEEs.
• To improve the identification of significant Single Nucleotide Polymorphisms (SNPs) while optimizing memory usage.
• To maintain stringent privacy guarantees for sensitive genotype data during distributed analysis.

Main Methods:

• Utilized a public training dataset to pre-identify significant SNPs for GWAS.
• Assigned dedicated memory to these identified SNPs for precise selection across the entire dataset.
• Integrated a learning-augmented approach into the SkSES framework to enhance SNP identification accuracy.
• Ensured sensitive genotype data remains undisclosed, upholding privacy in TEEs.

Main Results:

• The learning-augmented SkSES achieved up to 40% higher accuracy compared to the original SkSES.
• The method demonstrated improved scalability and effectiveness for collaborative GWAS.
• Optimized memory usage while enhancing the precision of significant SNP selection.

Conclusions:

• The learning-augmented SkSES significantly improves the accuracy and efficiency of privacy-preserving GWAS in TEEs.
• This advancement enhances the feasibility of large-scale, collaborative genomic studies.
• The method successfully balances computational efficiency, accuracy, and data privacy in TEEs.