Relatedness and Genotype × Environment Interaction Affect Prediction Accuracies in Genomic Selection: A Study in Cassava

Abstract

T he revolution in sequencing technologies has enabled fast and relatively inexpensive genome information (Metzker, 2010 Abbreviations: AMMI, additive main effect and multiplicative interaction; a top10 , mean relatedness of the top10 individuals in the validation set to those in the training set; AYT, advanced yield trial; BLUE, best linear unbiased estimator; BLUP, best linear unbiased predictor; CMD, cassava mosaic disease; CMDI, cassava mosaic disease incidence; CV-CR, cross-validation close relatives; CV-GE, crossvalidation genotype × environment; CV-Random, random crossvalidation; CV-Random_Half, cross-validation scheme in which a randomly chosen half of the observations are used; CV-noCR, cross-validation no close relatives; DM, root dry matter content; GS, genomic selection; G×E, genotype × environment; GBS, genotyping by sequencing; MAS, marker-assisted selection; MCBBI, mean cassava bacterial blight incidence; PYT, preliminary yield trial; RCBD, randomized complete-block design; RKHS, reproducing kernel Hilbert spaces; SNP, single nucleotide polymorphism; top10, 10 most closely related individuals; UYT, uniform yield trial.

Marker-based estimates reveal significant non-additive effects in clonally propagated cassava (Manihot esculenta): implications for the prediction of total genetic value and the selection of varieties

“…The models we fit were similar to those described in Ly et al (2013). The full model was specified as follows:…”

“…Efforts to improve cassava genetically with cutting edge methodologies including transgenic and genomic selection (GS) approaches are underway thanks to new genomic resources (Prochnik et al 2012;ICGMC 2015). Prediction with additive models has recently been evaluated (Oliveira et al 2012;Ly et al 2013) and genomic selection using standard models is currently being tested (http://www.nextgencassava.org). Vegetatively propagated crop (e.g., cassava) breeding can exploit nonadditive genetic effects by identifying superior clones as varieties (Ceballos et al 2015).…”

“…If the limited number of parents tested thus far represents the broader cassava breeding germplasm, genetic gains, especially for already lowheritability root yield traits, will be slow regardless of the breeding scheme employed (e.g., phenotypic vs. pedigree vs. genomic selection). Breeding gains have indeed been slow in cassava (Ceballos et al 2012) and low accuracies have been reported for genomic prediction of yield compared with cassava mosaic disease (CMD) resistance and dry matter (DM) content (Oliveira et al 2012;Ly et al 2013). However, cassava varieties are evaluated and disseminated to farmers by clonal propagation, meaning that accurate prediction of total (additive plus nonadditive) genetic value could contribute to variety selection.…”

Perspectives on the Application of Next-generation Sequencing to the Improvement of Africa’s Staple Food Crops

“…Given the long cycle of breeding, African staple crops such as cassava are set to benefit from GS approaches [117,118,120], where preliminary results have indicated reduced time of breeding cycle and reasonable prediction accuracy in some traits. Various ways of refining the prediction models via repeated phenotypic evaluations are being considered.…”

“…These approaches are being adopted in crops of importance in developing countries such as in maize and wheat [121], rice [124], pulses (legumes) [11], cassava [118,120], cowpea [125], lentil [126], soybean [127,128], and pigeon pea [129]. With respect to the best practice for GS, various models are being put forward [113].…”