Breeding with Major and Minor Genes: Genomic Selection for Quantitative Disease Resistance

Authors: MERRICK, LANCE - Washington State University; BURKE, ADRIENNE - Washington State University; Chen, Xianming; CARTER, AARON - Washington State University

Submitted to: Frontiers in Plant Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/6/2021
Publication Date: 8/6/2021
Citation: Merrick, L.F., Burke, A.B., Chen, X., Carter, A.H. 2021. Breeding with Major and Minor Genes: Genomic Selection for Quantitative Disease Resistance. Frontiers in Plant Science. 12. Article 713667. https://doi.org/10.3389/fpls.2021.713667.
DOI: https://doi.org/10.3389/fpls.2021.713667

Interpretive Summary: Most disease resistance in plants is quantitative, with both major and minor genes controlling resistance. This research aimed to optimize genomic selection (GS) models for use in breeding programs needing to select both major and minor genes for resistance. In this experiment, stripe rust of wheat was used as a model for quantitative disease resistance. The quantitative nature of stripe rust is usually phenotyped with two disease traits, infection type and disease severity. We compared two types of training populations composed of 2,630 breeding lines phenotyped in single plot trials from four years (2016-2020) and 475 diversity panel lines from four years (2013-2016), both across two locations. We also compared the accuracy of models with four different major gene markers and genome-wide association (GWAS) markers as fixed effects. The prediction models used 31,975 markers replicated 50 times using 5-fold cross-validation. We then compared the GS models with marker-assisted selection to compare the prediction accuracy of the markers alone and in combination. The GS models had higher accuracies than marker-assisted selection and reached an accuracy of 0.72 for disease severity. The major gene and GWAS markers had only a small to zero increase in prediction accuracy over the base GS model, with the highest accuracy increase of 0.03 for major markers and 0.06 for GWAS markers. There was a statistical increase in accuracy by using the disease severity trait, the breeding lines, population type, and by combing years. There was also a statistical increase in accuracy using major markers within the validation sets as the mean accuracy decreased. The inclusion of fixed effects in low prediction scenarios increased accuracy up to 0.06 for GS models using significant GWAS markers. Our results indicate that GS can accurately predict quantitative disease resistance in the presence of major and minor genes.

Technical Abstract: Most disease resistance in plants is quantitative, with both major and minor genes controlling resistance. This research aimed to optimize genomic selection (GS) models for use in breeding programs needing to select both major and minor genes for resistance. In this experiment, stripe rust (Puccinia striiformis Westend. f. sp. tritici Erikss.) of wheat (Triticum aestivum L.) was used as a model for quantitative disease resistance. The quantitative nature of stripe rust is usually phenotyped with two disease traits, infection type and disease severity. We compared two types of training populations composed of 2,630 breeding lines phenotyped in single plot trials from four years (2016-2020) and 475 diversity panel lines from four years (2013-2016), both across two locations. We also compared the accuracy of models with four different major gene markers and genome-wide association (GWAS) markers as fixed effects. The prediction models used 31,975 markers replicated 50 times using 5-fold cross-validation. We then compared the GS models with marker-assisted selection to compare the prediction accuracy of the markers alone and in combination. The GS models had higher accuracies than marker-assisted selection and reached an accuracy of 0.72 for disease severity. The major gene and GWAS markers had only a small to zero increase in prediction accuracy over the base GS model, with the highest accuracy increase of 0.03 for major markers and 0.06 for GWAS markers. There was a statistical increase in accuracy by using the disease severity trait, the breeding lines, population type, and by combing years. There was also a statistical increase in accuracy using major markers within the validation sets as the mean accuracy decreased. The inclusion of fixed effects in low prediction scenarios increased accuracy up to 0.06 for GS models using significant GWAS markers. Our results indicate that GS can accurately predict quantitative disease resistance in the presence of major and minor genes.