Fig. 4
From: COLLAGENE enables privacy-aware federated and collaborative genomic data analysis
GWAS p-value concordance and time/memory requirements. a The p-value concordance between secure federated GWAS (x-axis) and plain federated GWAS (y-axis) using the simulated genotype data among 3 sites. The Spearman R2 of the correlation is reported at the top of each plot. b The p-value concordance between federated GWAS (x-axis) and plink2 GWAS using pooled simulated dataset (y-axis). c The time usage of secure federated GWAS (in seconds) at each step of null model training (x-axis) and at each epoch (bar colors). d The time usage of secure GWAS for p-value assignment. e Peak memory usage (in gigabytes) of secure null model fitting. f Peak memory usage of secure p-value assignment. g Difference between total time usage (in seconds) of GWAS protocol using base and optimized build of COLLAGENE. Each bar shows the total time usage difference between base and optimized builds for one step in the protocol. For steps 1–8 (null model fitting with multiple epochs), time usage of the protocol is summed over all epochs. Blue (red) bars indicate that optimized build requires less (more) time than base build. h Difference between peak memory usage (in megabytes) for base and optimized build of COLLAGENE. Each bar shows the difference in memory usage for one step in the protocol. Red bars indicate that optimized build uses slightly more memory