Multi-trait genomic prediction using in-season physiological parameters increases prediction accuracy of complex traits in US wheat

Abstract

Background Recently genomic selection (GS) has emerged as an important tool for plant breeders to select superior genotypes. Multi-trait (MT) prediction model provides an opportunity to improve the predictive ability of expensive and labor-intensive traits. In this study, we assessed the potential use of a MT genomic prediction model by incorporating two physiological traits (canopy temperature, CT and normalized difference vegetation index, NDVI) to predict 5 complex primary traits (harvest index, HI; grain yield, GY; grain number, GN; spike partitioning index, SPI; fruiting efiiciency, FE) using two cross-validation schemes CV1 and CV2. Results In this study, we evaluated 236 wheat genotypes in two locations in 2 years. The wheat genotypes were genotyped with genotyping by sequencing approach which generated 27,466 SNPs. MT-CV2 (multi-trait cross validation 2) model improved predictive ability by 4.8 to 138.5% compared to ST-CV1(single-trait cross validation 1). However, the predictive ability of MT-CV1 was not significantly different compared to the ST-CV1 model. Conclusions The study showed that the genomic prediction of complex traits such as HI, GN, and GY can be improved when correlated secondary traits (cheaper and easier phenotyping) are used. MT genomic selection could accelerate breeding cycles and improve genetic gain for complex traits in wheat and other crops.

(Quasi) multitask support vector regression with heuristic hyperparameter optimization for whole-genome prediction of complex traits: a case study with carcass traits in broilers

“…As observed in other studies ( Aroujju et al 2020 ; Shahi et al . 2022 ), the MT methods generally did not present a better performance than ST did when observations for the correlated indicator trait were omitted in the testing set (CV1).…”

Genomic selection strategies to increase genetic gain in tea breeding programs

“…Instead, data from professional testers and biochemical traits from analytical techniques in the late stages can be used to train the GS model in the nursery stage. Genotyped seedlings could be selected for multiple traits such as tea quality and yield simultaneously using a multi‐trait GS model framework and a selection index (Shahi et al., 2022). Selection for better tea quality could thus begin at the earliest stage, where variation in quality traits is greater.…”

Multi-trait ridge regression BLUP with de novo GWAS improves genomic prediction for haploid induction ability and agronomic traits of haploid inducers in maize

“…The PA of the single-trait rrBLUP method of DTF, HIR, and PHT in MAGIC DHIs was moderate to high (Table 3), while employing multiple-trait GP models can increase PA even further for MAGIC DHIs. Using multitrait as multiple response variables in the GP models, the PA of low-heritability traits (<0.2) can take advantage of leveraging the genetic correlation with high-heritability traits (Jia and Jannink, 2012;Gaire et al, 2022;Shahi et al, 2022). In addition, synthetic populations generated by less than eight parents can yield high PA for GS (Schopp et al, 2017).…”

“…However, lines in breeding programs are usually evaluated for multiple traits. Simulation and empirical studies revealed that implementing multi-trait response variables in GP models has a higher prediction ability than single-trait GP models by leveraging estimates of covariance among genetically correlated traits (Jia and Jannink, 2012;Bhatta et al, 2020;Gill et al, 2021;Cappa et al, 2022;Shahi et al, 2022). The benefits of using multi-trait models are more pronounced for traits with low heritability and phenotypic traits that require costly measurements (Guo et al, 2014;Gaire et al, 2022).…”