Resource allocation optimization with multi-trait genomic prediction for bread wheat (Triticum aestivum L.) baking quality

Abstract

Key MessageMulti-trait genomic prediction models are useful to allocate available resources in breeding programs by targeted phenotyping of correlated traits when predicting expensive and labor-intensive quality parameters.AbstractMulti-trait genomic prediction models can be used to predict labor-intensive or expensive correlated traits where phenotyping depth of correlated traits could be larger than phenotyping depth of targeted traits, reducing resources and improving prediction accuracy. This is particularly important in the context of allocating phenotyping resource in plant breeding programs. The objective of this work was to evaluate multi-trait models predictive ability with different depth of phenotypic information from correlated traits. We evaluated 495 wheat advanced breeding lines for eight baking quality traits which were genotyped with genotyping-by-sequencing. Through different approaches for cross-validation, we evaluated the predictive ability of a single-trait model and a multi-trait model. Moreover, we evaluated different sizes of the training population (from 50 to 396 individuals) for the trait of interest, different depth of phenotypic information for correlated traits (50 and 100%) and the number of correlated traits to be used (one to three). There was no loss in the predictive ability by reducing the training population up to a 30% (149 individuals) when using correlated traits. A multi-trait model with one highly correlated trait phenotyped for both the training and testing sets was the best model considering phenotyping resources and the gain in predictive ability. The inclusion of correlated traits in the training and testing lines is a strategic approach to replace phenotyping of labor-intensive and high cost traits in a breeding program.Electronic supplementary materialThe online version of this article (10.1007/s00122-018-3186-3) contains supplementary material, which is available to authorized users.

Genomic selection in multi‐environment plant breeding trials using a factor analytic linear mixed model

“…In transferring GS to plants, however, many applications have failed to consider the inherent structure of plant breeding MET data sets. A recent example demonstrated GS in bread baking quality traits for a large number of varieties and environments (Lado et al., ). They used a stagewise approach where the adjusted means from the first stage were incorporated into a correlation matrix between environments.…”

“…The approach also fails to account for non‐genetic sources of variation in the field and laboratory phases nor does it appropriately model the variety (and hence marker) by environment interaction. Similar stagewise methods with deficient models for VEI are prolific in recent literature (see for example, Ben Hassen, Bartholomé, Valè, Cao, & Ahmadi, ; He et al., ; Lado et al., ; Michel et al., ).…”

Multi-trait regressor stacking increased genomic prediction accuracy of sorghum grain composition

“…size, genetic relatedness between individuals in training and testing population, marker density, span 15 of linkage disequilibrium and genetic architecture of traits are some of the factors that can affect the 16 predictive ability of the models [8][9][10]. Genomic prediction models are routinely studied and applied 17 by breeding programs around the world in several crops. Novel statistical methods that are capable 18 of incorporating pedigree, genomic, and environmental covariates into statistical-genetic prediction 19 models have emerged as a result of extensive computational research [11].…”

“…To address those challenges, 32 expanded genomic prediction models that perform joint analysis of multiple traits have been studied 33 using empirical and simulated data [15,16]. Subsequent improvement in prediction accuracy from 34 multi-trait model over single-trait model depends on trait heritability and correlation between the 35 traits involved [15,17]. 36 Data generated in breeding programs span multiple environment and are recorded for multiple 37 traits for each individual.…”

“…[18]. The use of multi-trait models has been consistently shown to increase prediction accuracy 302 over single-trait models across different crops and traits [15][16][17]49]. The multi-target regressor 303 stacking increased average prediction accuracy by 41% and 32% in the GSDP and RILs, respectively, 304 as compared to the STSE prediction accuracy.…”