Assessment of the performance of hidden Markov models for imputation in animal breeding

Table 2 Run time and accuracy for phasing, and haploid imputation methods on the real dataset scenario

Phasing method	Imputation method	Computing time (h)				Accuracy
Phasing method	Imputation method	HD phasing	LD phasing	Imputation	Total	Accuracy
HAPI-UR	IMPUTE2	11.53	43.09	5.63	60.25	0.997
HAPI-UR	Minimac3	11.53	43.09	2.27	56.89	0.995
HAPI-UR	Beagle v4.1	11.53	43.09	2.69	57.31	0.939
Eagle2	IMPUTE2	4.48 (8 cores)	2.37 (8 cores)	5.63	12.48	0.827
Eagle2	Minimac3	4.48 (8 cores)	2.37 (8 cores)	2.21	9.06	0.992
Eagle2	Beagle v4.1	4.48 (8 cores)	2.37 (8 cores)	4.19	11.04	0.925

The run time is given in hours separately for phasing and imputation steps and as a total. For Eagle2, the program was run distributed across 8 compute cores. HAPI-UR was run on a single core
HD high-density, LD low-density

ISSN: 1297-9686