Projects per year
The increasing global food demand in the context of climate change is a major concern of the 21st century. Developing crop varieties with improved adaptability to variable environmental conditions might be crucial to ensure food supply. Sorghum [Sorghum bicolor (L.) Moench] is a staple cereal crop in semi-arid regions of the world and is also grown worldwide as feed and bioenergy crop. Its drought-tolerance ability makes it a strategic crop for sustainable grain production. Modern approaches to quantitative genetics and statistical models such as genomic-assisted breeding techniques offer new opportunities for further improvements in crop productivity and adaptability. Compared to other major cereal crops such as maize and wheat, the application of these approaches remains largely unexplored for grain sorghum. The motivation of this thesis is to implement statistical models that exploit information from phenotypic, pedigree and genomic data for improving genetic evaluation and selection in sorghum breeding.
Chapter 2 assessed the performance of a novel method for spatial analysis of plant breeding field experiments based on two-dimensional smoothing with P-splines. This method was evaluated in comparison with the conventional spatial models by considering the improvement in precision and predictions of genetic effects in early generation sorghum breeding trials. The Chapter shows that both spatial methods produced equivalent performance. Differences in model parameterization as well as the advantages of the new spatial approach for routine application are discussed.
In Chapter 3, the impact of using pedigree and genomic information on prediction quality was explored for different traits in sorghum. For this, the Chapter proposes to use BLUP models fitting weighted combinations of pedigree and genomic relationship matrices, where the best-predictive combination is identified empirically in each prediction scenario. Results showed that the use of a merged pedigree–genomic matrix always improved predictive ability and unbiasedness of prediction relative to conventional G-BLUP, mainly for the traits with lower heritabilities. Based on these outcomes, the inclusion of pedigree information in genomic models is recommended to optimize predictions when the additive variation is not fully explained by markers.
Chapter 4 presents an extension of the study in Chapter 3 to the context of multi-trait genomic prediction. Specifically, we assessed the capacity of multi-trait models to improve genomic prediction for grain yield and stay-green in sorghum by using information from correlated auxiliary traits. In general, results showed that genomic prediction for both target traits can be enhanced by combining information from specific sets of traits. Predictions from conventional multi-trait G-BLUP were also optimized by combining pedigree and genomic information.
Chapter 5 investigated the effect of modelling genotype-by-environment interaction (G×E) on genomic prediction for grain yield in drought-stress and non-stress environments. Results indicated that accommodating G×E in genomic models was beneficial to improve the quality of prediction for specific adaptation as well as for broad adaptation to both types of environments. This Chapter also tested if better genomic predictions can be obtained by accounting for heterogeneous variances of marker effects. We found that weighting individual markers based on estimated locus-specific variances produced important improvements in predictive performance of genomic models, even for a largely polygenic trait such as grain yield.
To conclude, this thesis deals with different challenging aspects that may affect genetic evaluation in modern sorghum breeding. Specifically, several statistical modeling strategies making use of different sources of information have been proposed and assessed. The findings presented in this thesis are expected to contribute to increase the efficiency of selection schemes not only for sorghum but for crops in general.
|Qualification||Doctor of Philosophy|
|Award date||17 Mar 2020|
|Place of Publication||Wageningen|
|Publication status||Published - 2020|
FingerprintDive into the research topics of 'Statistical modeling of phenotypic, pedigree and genomic information for improved genetic evaluation in modern plant breeding: a case study with sorghum'. Together they form a unique fingerprint.
- 1 Finished
Advanced mixed models for plant breeding applications.
Velazco, J., Engel, B., Malosetti Zunin, M., van Eeuwijk, F., Velazco, J., van Eeuwijk, F. & Malosetti Zunin, M.
25/03/15 → 17/03/20