@PHDTHESIS{ 2021:441295443,
 	title = {USE OF SIMULATIONS TO OPTIMIZE SPATIAL MODELING AND LONG-TERM GENETIC GAINS IN SOYBEAN IN THE POST-GENOMIC ERA},
 	year = {2021},
 	url = "http://tede.unicentro.br:8080/jspui/handle/jspui/2278",
 	abstract = "This study covers two main topics: (1) Genomic assisted breeding has become an 
 essential tool in soybean breeding. However, the impact of different genomic selection 
 approaches on the short-term, and long-term gain is not well understood. To understanding 
 better this, a simulation with 200 breeding cycles was performed with a combination of family’s
 sizes, genomic prediction models (BayesA, BayesB, FLM, GBLUP, Random Forest,
 Phenotype, True Breeding Value), breeding strategies (AC: across families, WPSF: within and 
 across families, WF: within families) and selection intensities (2.5,5,7.5,10%). Selections were
 performed at the F4 generation, where individuals were phenotyped and genotyped. WF 
 selections provided the most significant long-term genetic gains. GBLUP and Bayesian 
 methods outperformed random forest and provided most of the genetic gains within the first 
 100 generations, outperformed by phenotypic selection after 100º generation. All methods 
 provided similar performance under WPSF selections. Faster decay in genetic variance was 
 observed when individuals were selected AC and WPSF, as 80% of the genetic variance was 
 reduced within 28 to 58 cycles. In contrast, WF selections preserved the variance up to 184º 
 cycle. Surprisingly, the selection intensity had less impact on long-term gains than breeding 
 strategies. The study supports those genetic gains can be optimized in the long term with 
 specific combinations of prediction models, families size, selection strategies, and selection 
 intensities. (2) Modeling field spatial patterns are standard practice for analyzing plant breeding 
 experimental trials to improve the accuracy of estimated of genetic parameters. This study aims 
 to quantify the accuracy and bias of estimated parameters using different approaches. We 
 contrasted three settings for the genetic term: no relationship (I), pedigree relationship (A), and 
 relationship based on molecular markers (G); and a set of approaches for the spatial variation: 
 no-spatial (NS), moving average covariate (MA), row-column adjustment (RC), autoregressive 
 AR1×AR1 (AR), spatial stochastic partial differential equations, or SPDE (SD), nearest 
 neighbor graph (NG), and Gaussian kernel (GK). Simulations were set to represent soybean 
 field trials at F2:4 generation, where the individuals were genotyped at high-density. The 
 simulated residual-to-signal ratio between residual variance and spatial variance (Ve:Vs)
 ranged from 9:1 to 1:9. Experimental settings were conducted under an augmented block
 design. Relationship information had a substantial impact on the accuracy of estimated of the
 genetic values (G>A>I) and contributed to estimated of the accuracy of spatial effects (30.63–
 42.27%). Negligible differences in accuracy of estimated of genetic parameters were observed 
 among models with different spatial terms when the spatial variation was equal to or lower than 
 a residual variance. The average improvement attributed to adequate spatial adjustment across 
 scenarios was 7-9%. Spatial models were ranked based on an improvement in the accuracy of 
 estimated of genetic parameters as SD ≥ GK ≥ AR ≥ NG ≥ MA > RC ≥ NS and the spatial 
 accuracy as GK ≥ SD ≥ NG > AR ≥ MA > RC. Estimates of genetic and spatial variance were 
 generally biased downwards, whereas residual variances were biased upwards. Both studies 
 have shown that statistical modeling in breeding programs broadly and jointly is beneficial in 
 selecting the best genotypes.",
 	publisher = {Universidade Estadual do Centro-Oeste},
 	scholl = {Programa de Pós-Graduação em Agronomia (Doutorado)},
 	note = {Unicentro::Departamento de Agronomia}
}