pubmed.ncbi.nlm.nih.gov

An experimental validation of genomic selection in octoploid strawberry - PubMed

  • ️Sun Jan 01 2017

An experimental validation of genomic selection in octoploid strawberry

Salvador A Gezan et al. Hortic Res. 2017.

Abstract

The primary goal of genomic selection is to increase genetic gains for complex traits by predicting performance of individuals for which phenotypic data are not available. The objective of this study was to experimentally evaluate the potential of genomic selection in strawberry breeding and to define a strategy for its implementation. Four clonally replicated field trials, two in each of 2 years comprised of a total of 1628 individuals, were established in 2013-2014 and 2014-2015. Five complex yield and fruit quality traits with moderate to low heritability were assessed in each trial. High-density genotyping was performed with the Affymetrix Axiom IStraw90 single-nucleotide polymorphism array, and 17 479 polymorphic markers were chosen for analysis. Several methods were compared, including Genomic BLUP, Bayes B, Bayes C, Bayesian LASSO Regression, Bayesian Ridge Regression and Reproducing Kernel Hilbert Spaces. Cross-validation within training populations resulted in higher values than for true validations across trials. For true validations, Bayes B gave the highest predictive abilities on average and also the highest selection efficiencies, particularly for yield traits that were the lowest heritability traits. Selection efficiencies using Bayes B for parent selection ranged from 74% for average fruit weight to 34% for early marketable yield. A breeding strategy is proposed in which advanced selection trials are utilized as training populations and in which genomic selection can reduce the breeding cycle from 3 to 2 years for a subset of untested parents based on their predicted genomic breeding values.

PubMed Disclaimer

Figures

Figure 1
Figure 1

Predictive ability (PA) for true validation of several genomic selection (GS) methods, based on a model fitted in either T1/2013 or T2/2013 to predict both T1/2014 and T2/2014. The mean corresponds to the average of GS methods. (a) model fitted in T1/2013 to predict T1/2014, (b) model fitted in T1/2013 to predict T2/2014, (c) model fitted in T2/2013 to predict T1/2014, and (d) model fitted in T2/2013 to predict T2/2014.

Figure 2
Figure 2

Scatter plot of pairwise regular linkage disequilibrium (LD) (r2) and LD corrected for relatedness (rv2) against increasing genetic distances between markers on the LG 6A. (a) pairwise regular LD (r2) for individuals tested in the T2/2013 trial, (b) LD corrected for relatedness (rv2) for the T2/2013 trial, (c) pairwise regular LD (r2) for the T2/2014 trial, and (d) LD corrected for relatedness (rv2) over increasing genetic distances of LG 6A for the T2/2014 trial. Green lines indicate smoothed splines.

Figure 3
Figure 3

Selection efficiency based on genetic gains when selecting the top 10% of genotypes (corresponding to 30 individuals) for three different GS models fitted with T2/2013 phenotypic and marker data for training and used to make predictions in T2/2014 using T2/2014 marker data only. Gains are compared with breeding values for T2/2014 estimated using both phenotypic and marker data from T2/2014. AWT, average weight (g per fruit); EMY, early marketable yield (g per plant); SSC, soluble solids content (%); TC, proportion of total culls (%); TMY, total marketable yield (g per plant).

Figure 4
Figure 4

Breeding strategy transition to an approach combining marker-assisted seedling selection (MASS) for major loci and genomic selection (GS) as an early parent selection method for complext traits. The incorporation of GS for parent selection allows the use of first-year seedlings in crosses prior to their inclusion in a replicated advanced selection trials (T2). Trials of unselected seedlings (T1) will be eliminated in favor of increasing the number of genotypes evaluated in T2 trials, which will be used as training populations for future GS models.

Similar articles

Cited by

References

    1. Meuwissen TH, Hayes BJ, Goddard ME. Prediction of total genetic value using genome-wide dense marker maps. Genetics 2001; 157: 1819–1829. - PMC - PubMed
    1. Henderson CR. Applications of Linear Models in Animal Breeding. University of Guelph: Guelph, Ontario, Canada, 1984.
    1. Heffner EL, Sorrels ME, Jannick JL. Genomic selection for crop improvement. Crop Sci 2009; 49: 1–12.
    1. Hayes BJ, Bowman PJ, Chamberlain AJ, Goddard ME. Invited review: Genomic selection in dairy cattle: progress and challenges. J Dairy Sci 2009; 92: 433–443. - PubMed
    1. Resende MFR Jr, Muñoz P, Resende MDV, Garrick DJ, Fernando RL, Davis JM et al. Accuracy of genomic selection methods in a standard dataset of Loblolly pine (Pinus taeda L.). Genetics 2012; 190: 1503–1510. - PMC - PubMed