Author : Nguyen Van Hieu

Summary : Rice (Oryza sativa) is the main source of calories for over half of the world's population. Rice diversity is large and has been shaped by natural selection and domestication. Part of it has been the basis of breeding programs for creating modem rice varieties. The efforts to select high-yielding cultivars during the past decades have led to a reduction of the genetic diversity used by breeders. With the integration of genomic tools and the pursuit of higher genetic gains, breeders need a better characterization of their elite pool to meet short- and long-term targets. This PhD research aimed to gain a deeper insight into the genetic diversity of elite germplasm, thus enabling a more efficient selection. The breeding program at the International Rice Research Institute (IRRI, Philippines) was used in this work for its rich history and global mandate, with a focus on South Asia and Eastern Africa. The first objective was to characterize the genetic structure of the elite germplasm at a fine scale. The second objective was to evaluate genomic selection strategies that integrate the information from elite lines.
A panel of 124 elite lines representing the diversity of the material presently used by the program served as the basis of the analyses conducted in this research work. For the first objective, the elite lines were resequenced at medium depth to capture the allelic diversity. We found 360 regions ranging from 100kb to 1.42Mb with low diversity. These regions, which represent one-fifth of the genome, include genes that have proven agronomical impact and therefore deserve specific attention for managing diversity in the long term. In addition, we compared the elite lines with 254 accessions that represented the diversity of O. sativa and we could observe the general similarity of elite lines with the Xian/Indica group, and we revealed 83 genomic regions that differentiate between the two groups, which may bear adaptive genetic factors. For the second objective, the elite breeding lines were phenotyped across 15 environments in Asia and Africa. Seven multi­environment models were used to assess the efficiency of genomic prediction to predict untested environments. The predictive abilities for days to flowering were found to range from 0.06 to 0.79, for plant height from 0.25 to 0.88, and for grain yield from -0.29 to 0.62. We found that models integrating genotype-by-environment interaction effects did not significantly perform better than models integrating only main effects.
This work was complemented by a participation in exploration of exotic diversity through the characterization of peri-Himalayan germplasm with original adaptive traits.
These results will help in managing breeding program diversity and gaining deeper insights into the genetic factors that are critical to crop adaptation and performance. Furthermore, they will help refine the testing strategy, leading to improved accuracy and predictability of the genomic prediction models.