Table 3:

Runtime metrics of the Association Testing Module of BIGwas (regression analysis with PLINK and mixed model analysis with SAIGE) compared to association testing module of the H3Agwas Pipeline Version 3 (regression analysis with PLINK) [12] for several different-sized GWAS data sets

Sample No.	Variant No.	BIGwas Runtime	H3A Runtime
10,000	250,000	21  min	8  min
10,000	700,078	54  min	16  min
20,554	700,078	1  h 33  min	1  h 23  min
5,480	81,708,012	14  h 26  min	*
487,409	92,775,302	4  d 18  h	*
974,818	92,775,302	9  d 14  h	*

An imputed genome-wide input data set with almost 1 million samples and >92 million genetic variants (i.e., twice the size of the imputed UKB GWAS data set) can be tested for association within 10 days with 1 command of the BIGwas software, whereby only 150 jobs (configurable; equivalent to ∼7 compute nodes on our HPC cluster system) are used in parallel. *H3A uses PLINK genotype files as input (but not imputed allele dosages).