idMEGA

idMEGA performs federated genome-wide association studies by enabling association testing across multiple research sites while preserving genotype and phenotype confidentiality and correcting for population stratification.

Key Features:

• Federated Association Testing: Enables association testing by sharing intermediate statistics through a central server rather than raw genotype or phenotype data.
• Generalized Linear Mixed Models (GLMM): Employs generalized linear mixed models to model fixed and random effects for complex trait association testing.
• Population Stratification Correction: Uses a reference projection technique to correct for population stratification across study sites.
• Local-Gradient Updates: Implements efficient local-gradient updates among participating sites to improve computational efficiency and scalability.
• Privacy Protection: Avoids explicit sharing of sensitive genotype and phenotype data to maintain participant confidentiality.
• Confounding Factor Modeling: Addresses confounding factors, including population stratification and disease etiology, through combined use of GLMM and reference projection.

Scientific Applications:

• Large-scale collaborative GWAS: Integrates diverse datasets from multiple sites to identify genetic associations with complex traits without sharing raw data.
• Privacy-sensitive studies: Enables analysis of sensitive health or genotype-phenotype datasets under restrictive data protection constraints.
• Population-structured cohorts: Facilitates association testing in cohorts with population structure by applying reference projection correction.

Methodology:

Distributed computing where each site performs local computations and contributes intermediate statistics to a central server; federated association testing via shared intermediate statistics; generalized linear mixed models for fixed and random effects; reference projection for population stratification correction; and local-gradient updates among participating sites.