Full text
| |
| Author(s): |
Stephanie Karenina Bajay
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Biologia |
| Defense date: | 2022-06-14 |
| Examining board members: |
Anete Pereira de Souza;
Maria Imaculada Zucchi;
Prianda Rios Laborda;
Lorena Guimarães Batista;
Luís Felipe Ventorim Ferrão
|
| Advisor: | Anete Pereira de Souza |
| Abstract | |
The selection of superior genotypes, through the application of genomic selection (GS) and genome-wide association studies (GWAS), has dramatically improved the speed and scale of genetics applied to forest breeding. However, the choice of the best methodology to be adopted varies according to the aims to be achieved in each stage of development of the breeding program. GS is an excellent strategy to be adopted in the initial and intermediate steps of breeding programs, for the selection of superior genotypes. In recent years, efforts to improve perennial species have turned to GS, as it should allow tracking of most sources of variation for complex traits. In this context of GS, the incorporation of advanced methods of machine learning has also been highlighted, due to the fact that these algorithms allow training using more complex data representation and because they do not require assumptions about the model. Among the various existing learning algorithms, we applied in the present study, the feature selection, which aims to reduce the density of markers and build simple and comprehensive prediction models, avoiding the attribution of non-genetic effects to the markers and increasing the predictive power of the phenotypes. In addition to GS, with the objective of expanding the understanding of the architecture and genetic basis of the phenotypes studied, the strategy with GWAS was also adopted. For the interpretation of how the genes discovered by GWAS and GS influence the analyzed traits, strategies of annotation of genetic pathways and ontologies identified with transcriptomes were adopted. Gene co-expression networks were constructed in order to develop a global understanding of gene expression and biological function possibly correlated with candidate genes for modulating the phenotypes of interest. The combination of different statistics and genomic analyses, such as SG, machine learning, GWAS and gene co-expression networks, becomes a promising strategy to effectively deal with the improvement of complex traits. Thus, the main objective of this thesis was to integrate multiple omic analyses: SG, GWAS, ML and gene co-expression networks for the selection of superior genotypes in forest trees (AU) | |
| FAPESP's process: | 19/13452-3 - Genomic prediction and genome-wide association in rubber tree, main renewable source of rubber production |
| Grantee: | Stephanie Karenina Bajay |
| Support Opportunities: | Scholarships in Brazil - Doctorate |