Computational strategies for the preconditioned conjugate gradient method applied to ssSNPBLUP, with an application to a multivariate maternal model

Table 2 Wall clock times for the preparation and solving processes of different ssSNPBLUP systems

Wall clock time (s)	ssSNPBLUP_Liu (Plink)^a	ssSNPBLUP_Liu (DP)^b	ssSNPBLUP_MS^c
\(diag \left( {\mathbf {A}}_{gg}^{-1} \right)\)^d	136.69	139.78	136.20
Preconditioner^d	546.41	581.42	1,177.26
\({\mathbf {A}}^{-1}\)^d	50.62	57.40	50.77
\({\mathbf {Z}}{\mathbf {v}}\)^e	3.47	7.77	3.42
\(\mathbf {Z'}{\mathbf {v}}\)^e	1.53	4.61	1.45
Average time/iteration	12.82	20.23	16.89
Iterative process	43,074.48	67,961.84	107,041.71
Software total time	44,531.00	69,593.09	109,126.07

^assSNPBLUP model proposed by Liu et al. [9] and using the Plink 1 binary form; or ^busing double precision reals; ^cssSNPBLUP model proposed by Mantysaari and Stranden [11] and using the Plink 1 binary form; ^dWall clock time needed for the computation of the mentioned matrix; ^eMultiplication of the centered genotype matrix, or its transpose, by an array

ISSN: 1297-9686