Ridge regression and deep learning models for genome-wide selection of complex traits in New Mexican Chile peppers

Table 3 Response to selection (R) for total yield per plant and yield components using whole marker dataset for a Bayesian ridge regression model. The top 20 selection candidates (genotypes) with the highest phenotypic value (PV) and genomic estimated breeding values (GEBV), corresponding to a selection intensity of 10%, were selected for phenotypic selection (PS) and genomic selection (GS), respectively. The top 10% selection candidates with the highest PV and GEBV from PS and GS were selected in PS + GS.

Trait	H^{2 1}	Mean population without selection	Mean population with selection			Selection differential, S²			Response to selection, R³			Percent change relative to PS, R			No. of selection candidates selected (%)
Trait	H^{2 1}	Mean population without selection	PS	GS	PS + GS	PS	GS	PS + GS	PS	GS	PS + GS	PS	GS	PS + GS	PS + GS
Mature green	0.58	0.34	0.35	0.35	0.36	0.01	0.01	0.02	0.01	0.01	0.01	-	0	0	11 (55%)
Mature red	0.32	0.50	0.69	0.57	0.71	0.19	0.07	0.21	0.06	0.02	0.07	-	-67%	17%	9 (45%)
Ten pod weight	0.88	0.36	0.77	0.76	0.8	0.41	0.4	0.44	0.36	0.35	0.39	-	-3%	8%	17 (85%)
Total yield per plant	0.20	0.82	1.11	1.02	1.21	0.29	0.20	0.39	0.06	0.04	0.08	-	-33%	33%	9 (45%)

¹ Broad-sense heritability (H²) values reported by Lozada et al. [30]
² Selection differential, S = µ_{pop. with selection} - µ_{pop. without selection}
³ Response to selection, R = H²S

ISSN: 2730-6844