Genome wide selection methods applied to high biomass sorghum for the production of second generation ethanol

Author(s)

Amanda Avelar de Oliveira

• Document type: Master's Dissertation
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2015-07-03

Examining board members

Gabriel Rodrigues Alves Margarido; Cynthia Maria Borges Damasceno

• Advisor: Gabriel Rodrigues Alves Margarido

Abstract

Increased concerns about environmental issues have aroused global interest in the use of alternative fuels, and the use of plant biomass emerges as a viable alternative for the generation of biofuels. Different organic materials have been used, including high biomass sorghum (Sorghum bicolor L. Moench). Genomic selection has great potential and could, in the medium term, restructure plant breeding programs, promoting greater genetic gains when compared to other methods and significantly reducing the time required for the development of new cultivars through early selection. This work aimed at evaluating models of genomic selection and applying them to the prediction of breeding values for a panel of high biomass sorghum genotypes of Embrapa / Milho e Sorgo. This panel includes materials from the gene bank and materials used in sorghum breeding programs of this institution, as well as core collections from CIRAD and ICRISAT, and is therefore divided into two sub-panels. The 100 lines of sub-panel 1 were evaluated phenotypically for two years (2011 and 2012) and the 100 lines of sub-panel 2 for one year (2011), both in the city of Sete Lagoas, Minas Gerais, for the following phenotypic traits: days to flowering, plant height, fresh and dry matter yield and fiber, cellulose, hemicellulose and lignin proportions. Subsequently, the 200 lines were genotyped by via the genotyping by sequencing technique. From these genotypic and phenotypic data, genomic selection models Bayes A, Bayes B, Bayes Cπ, Bayes Lasso, Bayes Ridge Regression and Random Regression BLUP (RRBLUP) were fitted and compared. The predictive capabilities obtained were high and varied little between the different models, ranging from 0.61 for days to flowering to 0.85 for acid fiber, when the RRBLUP model was used on the combined analysis of the two sub-panels. On the other hand, cross prediction between sub-panels resulted in substantially lower predictive capability, never above 0.66 and in some scenarios virtually equal to zero, with greater variations between the fitted models. Simulations of using subsets of molecular markers are presented and indicate possibilities of achieving higher predictive capabilities. Functional enrichment analyses performed with the marker predicted effects suggested interesting associations, which should be investigated in more detail in future studies, with potential for elucidating the genetic architecture of quantitative traits. (AU)

• FAPESP's process: 13/25132-7 - Genome wide selection methods applied to high biomass sorghum for the production of second generation ethanol
• Grantee: Amanda Avelar de Oliveira
• Support Opportunities: Scholarships in Brazil - Master