Integration of Genomic Selection into the University of Florida Strawberry Breeding Program Luis F. Osorio, Salvador Gezan, Vance Whitaker, Sujeet Verma 8 th International Rosaceae Genomics Conference June 22,2016
Genomic Selection Process Training Population Genotyping & Phenotyping Candidate Population Genotyping Calculate GEBV Make Selections (From Heffner et al., (2009)
Objectives Explore different GS models for some of the most important traits Use true validation for prediction Define a breeding strategy that incorporates GS
Breeding Cycle 2014 2015 2016 2017 Cross Cross Cross Cross Seedlings T1 Adv. Sel.
Breeding Strategy 2013 2014 2015 2016 Cross Cross Cross Cross Seedlings Seedlings Seedlings Seedlings T1 T1 T1 T1 Adv. Sel. Adv. Sel. Adv. Sel. Adv. Sel.
Experimental Design and Plant Material Two trials per season (T1, ) - RCBD Replicated seedling trials (T1) ~ 80 fam & ~ 580 genotypes 3 replicates/gen Replicated advanced selection trials () 300 genotypes 5 replicates/gen
Phenotyping Measurements (per trial/season) Traits No. Units Marketable Yield (TMY) 15 grams Average Fruit Wt (AWT) 15 grams Total Culls (TC) 15 % Brix (SSC) 5 %
Genotyping Marker Quality Control 17,479 markers were used in GS Markers with MAF < 5% were removed 90K IStraw90 Axiom SNP array Affimetrix & RosBREED (Bassil and Davis et al. 2015) Markers with missing values > 5% were removed Missing marker data was imputed by using the average allele frequency
Individual mixed models : Statistical Analyses and GS Models To estimate variance components for heritability estimates (ASReml Gilmour et al 2009) Phenotypic response to be used in the GS analyses G-BLUP: assumes marker effects have similar variances (Genomatrix -Nazarian and Gezan 2016 and ASReml-R) Bayes B and Bayes C: markers have different variance effects (BGLR - Perez and de los Campos, 2014) Reproducing kernel Hilbert spaces (RKHS): captures some non-additive effects (BGLR - Perez and de los Campos, 2014)
Comparison of GS Models - Criteria- Predictive Ability: Prediction Accuracy:,,, PBLUP: traditional pedigree-based estimates Efficiency of Parental Selection (5%, 10%): /
Comparison of GS Methods -True Validation- 2013 2014 2015 2016 Cross Cross Cross Cross Seedlings Seedlings Seedlings Seedlings T1 T1 T1 T1 Adv. Sel. Adv. Sel. Adv. Sel. Adv. Sel. GS
Heritabilities Trial Trait Parameter 2013-2014- AVW h² 0.46 0.25 H² 0.62 0.59 SSC h² 0.18 0.41 H² 0.34 0.46 TMY h² 0.11 0.26 H² 0.38 0.47 TC h² 0.44 0.08 H² 0.47 0.53
Predictive Ability Trait PBLUP GS Models GBLUP Bayes B Bayes C RKHS AWT 0.444 0.490 0.494 0.488 0.515 SSC 0.371 0.427 0.438 0.436 0.451 TMY 0.238 0.306 0.353 0.337 0.333 TC 0.139 0.320 0.350 0.352 0.318
Prediction Accuracy Trait PBLUP GS Models GBLUP Bayes B Bayes C RKHS AWT 0.549 0.606 0.610 0.603 0.636 SSC 0.630 0.725 0.744 0.740 0.766 TMY 0.507 0.652 0.753 0.718 0.710 TC 0.159 0.365 0.400 0.402 0.363
Parental Selection Efficiency 80.0% 70.0% 74.4% 71.1% 73.7% 70.4% Genetic Gain 60.0% 50.0% 40.0% 30.0% 20.0% 52.6% 51.7% 51.5% 48.4% 28.1% 38.8% 55.8% 52.9% 55.0% 49.9% 46.5% 45.8% 10.0% 0.0% AWT SCC TC TMY AWT SCC TC TMY GBLUP 5% 10% Bayes B
Conclusions Bayes B showed better predictive ability than other models Predictive ability is linearly related to the heritability of the trait Prediction accuracy was high (>0.60) for all models and traits except for TC The efficiency of parental selection was moderate to high (> 45%) for all traits
New Breeding Cycle 2016 2017 2018 2019 Cross Cross Cross Cross Seedlings X Adv. Sel.
New Breeding Cycle 2016 2017 2018 Cross Cross Cross 2019 Cross GS Validation Population Seedlings Training Population Adv. Sel. GS Adv. Sel.
New Breeding Strategy 2016 2017 2018 2019 Cross Cross Cross Cross GS GS Seedlings Seedlings Seedlings Seedlings T1 Adv. Sel. Adv. Sel. Adv. Sel. Adv. Sel. GS GS
GS Advantages for the UF Strawberry Breeding Program Reducing time, labor and expenses by eliminating T1 tests Accelerating the evaluation of potential parents and their use in crossing Increasing genetic gains by using selected parents earlier in the breeding cycle Increasing genetic gains by improving the accuracy of prediction of genetic values
Acknowledgements Strawberry Group-GCREC Angelita Arredondo Kelsey Cearley Jin Hee Kim Jose Hernandez Jozer Mangandi David Moore Catalina Moyer Young Hee Noh Seonghee Lee Natalia Salinas Zhen Fan SFRC (UF) Alireza Nazarian RosBREED Project Quantitative Genetics team
Acknowledgements - RosBREED Project-