Abstract:
This study addressed the privacy challenges in genome-wide association studies
(GWAS) by developing a secure system that performs computations directly on encrypted
genotype data using Fully Homomorphic Encryption (FHE). The system
was implemented with the CKKS scheme via the TenSEAL library and structured
using a Django client and Flask server. Five conditional logic algorithms
were evaluated—Polynomial Approximation, Multiplexer, Blind Evaluation, Conditional
Branching, and Minimax Approximation—to compute GWAS statistics
such as allelic odds ratio, chi-square, minor allele frequency, and Hardy-Weinberg
equilibrium. Results showed that Multiplexer and Conditional Branching achieved
the highest accuracy, while Polynomial and Minimax approaches offered trade-offs
in speed and flexibility. The system demonstrated that secure GWAS analysis is
feasible without compromising data privacy.
