SGX

SGX performs genotype imputation inside an Intel Software Guard Extensions (SGX) Trusted Execution Environment (TEE), inferring missing genotypes from reference genome panels using HMM-based algorithms to enable privacy-preserving analyses.

Key Features:

• Privacy-Preserving Imputation: Processes input genomes exclusively within an Intel SGX Trusted Execution Environment (TEE) to keep data confidential from the service provider.
• Hardware-Based Security: Leverages Intel SGX hardware enclaves to protect genomic data and computations from unauthorized access.
• Side-Channel Resilience: Implements measures to mitigate side-channel attacks identified in existing imputation software.
• Efficient Imputation Algorithm: Uses a hidden Markov model (HMM)-based strategy similar to Minimac to achieve imputation accuracies comparable to Minimac.
• Scalability: Designed to scale to large genomic datasets while maintaining the stated security and imputation accuracy.

Scientific Applications:

• Genome-wide association studies (GWAS): Enables imputation of missing genotypes to increase marker density and statistical power in GWAS.
• Population genetics: Supports population genetic analyses using imputed genotypes.
• Personalized medicine: Facilitates downstream analyses in personalized medicine where genotype completeness and privacy are required.

Methodology:

Genotype imputation is executed inside the Intel SGX TEE and employs HMM-based strategies similar to Minimac, with enclave isolation preventing extraction of sensitive data from the host.