1. Genomes to Fields 2014 Workshop G × E ANALYSIS Diego Jarquin Chicago, IL December 10

2. Variance components for common inbreds and hybrids Genomes to Fields Inbred – Pollen DAPInbred – Plant height Hybrid – Plant height Hybrid – Pollen DAP

3. Bayesian AMMI Analysis for Inbreds – Plant Height Genomes to Fields

4. Bayesian AMMI Analysis for Hybrids - Plant height

5. Genomes to Fields Genomic Prediction $ $ Model training Predict and select Molecular markers Phenotypic data Test candidates  Predict performance of hybrids using their whole-genome marker profiles.  Potential to save time and money  How? Using genotypes and phenotypes to train a model to predict how each genotype is expected to perform in each environment. 530 Hybrids 11 Field trials 42% of the potentials cells were tested

6. Genomes to Fields Similarities among environments based on environmental covariates Genomic Prediction

7. Cross-validation schemes Genomes to Fields Leave-one-out CV within trials - Baseline CV0 - Predicting new trials CV1 - Predicting new hybrids

8. Prediction accuracy Plant height Genomic prediction Genomes to Fields

9. Prediction accuracy Pollen DAP Genomic prediction Genomes to Fields

10. Observed vs predicted – Plant Height

11. Genomes to Fields Observed vs predicted – Pollen DAP

12. Genomes to Fields Final Remarks  With large phenotypic and genotypic sets, good predictions for new environments (CV0) can be expected (E+G model).  By including information on the performance of other hybrids tested in these environments (CV1), accuracy was improved up to 38% using E+G model.  Best results were obtained considering the interaction components (GE or GW), increasing prediction accuracy up to 75% compared to CV0.  These results highlight the importance of capturing GxE effects for predicting phenotype from genotype.