Multi-omics-data-assisted genomic feature markers preselection improves the accuracy of genomic prediction

Table 3 Prediction accuracies using SNPs preselection based on TWAS results (S_TWAS)

Model	P-value cutoffs^a	Prediction accuracy (Mean ± SE^b)
		Startle response		Starvation resistance
		Female	Male	Female	Male
GBLUP^d	All^c	0.208 ± 0.020	0.181 ± 0.022	0.272 ± 0.017	0.307 ± 0.015
	< 0.05	0.189 ± 0.022	0.118 ± 0.022	0.128 ± 0.017	0.196 ± 0.015
	< 0.001	0.029 ± 0.023	−0.008 ± 0.026	−0.007 ± 0.021	0.081 ± 0.022
	< 0.0001	0.005 ± 0.022	−0.042 ± 0.019	0.002 ± 0.020	0.024 ± 0.017
	< 0.00001	−0.001 ± 0.022	−0.042 ± 0.019	0.002 ± 0.020	0.004 ± 0.020
GFBLUP^e	< 0.05	0.176 ± 0.024	0.106 ± 0.024	0.196 ± 0.020	0.287 ± 0.017
	< 0.001	0.168 ± 0.022	0.140 ± 0.022	0.266 ± 0.018	0.291 ± 0.016
	< 0.0001	0.170 ± 0.019	0.137 ± 0.024	0.272 ± 0.017	0.288 ± 0.018
	< 0.00001	0.169 ± 0.019	0.137 ± 0.024	0.272 ± 0.017	0.296 ± 0.016

^aP-value cutoffs: using different P-value cutoffs to preselect genes based on the results of transcriptome-wide association study (TWAS), then extracted SNPs from whole genome sequencing (WGS) data according corresponding genomic positions of genes; ^bSE: Standard error; ^cAll: All SNPs of WGS data; ^dGBLUP: Genomic best linear unbiased prediction; ^eGFBLUP: A genomic feature best linear unbiased prediction

ISSN: 2049-1891