ABSTRACT
The availability of high-dimensional genomic data and advancements in genome-based prediction models (GP) have revolutionized and contributed to accelerated genetic gains in soybean breeding programs. GP-based sparse testing is a promising concept that allows increasing the testing capacity of genotypes in environments, of genotypes or environments at a fixed cost, or a substantial reduction of costs at a fixed testing capacity. This study represents the first attempt to implement GP-based sparse testing in soybeans by evaluating different training set compositions going from non-overlapped RILs until almost the other extreme of having same set of genotypes observed across environments for different training set sizes. A total of 1,755 recombinant inbred lines (RILs) tested in nine environments were used in this study. RILs were derived from 39 bi-parental populations of the Soybean Nested Association Mapping (NAM) project. The predictive abilities of various models and training set sizes and compositions were investigated. Training compositions included a range of ratios of overlapping (O-RILs) and non-overlapping (NO-RILs) RILs across environments, as well as a methodology to maximize or minimize the genetic diversity in a fixed-size sample. Reducing the training set size compromised predictive ability in most training set compositions. Overall, maximizing the genetic diversity within the training set and the inclusion of O-RILs increased prediction accuracy given a fixed training set size; however, the most complex model was less affected by these factors. More testing environments in the early stages of the breeding pipeline can provide a more comprehensive assessment of genotype stability and adaptation which are fundamental for the precise selection of superior genotypes adapted to a wide range of environments.
ABSTRACT
The availability of high-dimensional molecular markers has allowed plant breeding programs to maximize their efficiency through the genomic prediction of a phenotype of interest. Yield is a complex quantitative trait whose expression is sensitive to environmental stimuli. In this research, we investigated the potential of incorporating soil texture information and its interaction with molecular markers via covariance structures for enhancing predictive ability across breeding scenarios. A total of 797 soybean lines derived from 367 unique bi-parental populations were genotyped using the Illumina BARCSoySNP6K and tested for yield during 5 years in Tiptonville silt loam, Sharkey clay, and Malden fine sand environments. Four statistical models were considered, including the GBLUP model (M1), the reaction norm model (M2) including the interaction between molecular markers and the environment (G×E), an extended version of M2 that also includes soil type (S), and the interaction between soil type and molecular markers (G×S) (M3), and a parsimonious version of M3 which discards the G×E term (M4). Four cross-validation scenarios simulating progeny testing and line selection of tested-untested genotypes (TG, UG) in observed-unobserved environments [OE, UE] were implemented (CV2 [TG, OE], CV1 [UG, OE], CV0 [TG, UE], and CV00 [UG, UE]). Across environments, the addition of G×S interaction in M3 decreased the amount of variability captured by the environment (-30.4%) and residual (-39.2%) terms as compared to M1. Within environments, the G×S term in M3 reduced the variability captured by the residual term by 60 and 30% when compared to M1 and M2, respectively. M3 outperformed all the other models in CV2 (0.577), CV1 (0.480), and CV0 (0.488). In addition to the Pearson correlation, other measures were considered to assess predictive ability and these showed that the addition of soil texture seems to structure/dissect the environmental term revealing its components that could enhance or hinder the predictability of a model, especially in the most complex prediction scenario (CV00). Hence, the availability of soil texture information before the growing season could be used to optimize the efficiency of a breeding program by allowing the reconsideration of field experimental design, allocation of resources, reduction of preliminary trials, and shortening of the breeding cycle.
ABSTRACT
Genomic selection (GS) has proven to be an efficient tool for predicting crop-rank performance of untested genotypes; however, when the traits have intermediate optima (phenology stages), this implementation might not be the most convenient. GS might deliver high-rank correlations but incurring in serious bias. Days to heading (DTH) is a crucial development stage in rice for regional adaptability with a significant impact on yield potential. The objective of this research consisted in develop a novel method that accurately predicts time-related traits such as DTH in unobserved environments. For this, we propose an implementation that incorporates day length information (DL) in the prediction process for two relevant scenarios: CV0, predicting tested genotypes in unobserved environments (C method); and CV00, predicting untested genotypes in unobserved environments (CB method). The use of DL has advantages over weather data since it can be determined in advance just by knowing the location and planting date. The proposed methods showed that DL information significantly helps to improve the predictive ability of DTH in unobserved environments. Under CV0, the C method returned a root-mean-square error (RMSE) of 3.9 days, a Pearson correlation (PC) of 0.98 and the differences between the predicted and observed environmental means (EMD) ranged between -4.95 and 4.67 days. For CV00, the CB method returned an RMSE of 7.3 days, a PC of 0.93 and the EMD ranged between -6.4 and 4.1 days while the conventional GS implementation produced an RMSE of 18.1 days, a PC of 0.41 and the EMD ranged between -31.5 and 28.7 days.
