Advances in soybean breeding: use of genomic and phenotypic tools for resistance to the stink bug complex

• Autor(a) principal: Oliveira, Maiara de
• Data de Publicação: 2024
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/11/11137/tde-06112024-180656/
• topic: Associações genômicas; Fenotipagem de alto rendimento; Genomic associations; Genomic selection; High-throughput phenotyping; Insect resistance; Melhoramento de precisão; Precision breeding; Resistência a Insetos; Seleção Genômica

Soybean is a crop of great global importance, especially in Brazil, where it faces significant threats from pests such as stink bugs, which cause considerable productivity losses. Genetic resistance to stink bugs is the most efficient control strategy, but its quantitative nature makes implementation in breeding programs challenging. Traditionally, manual phenotyping is labor-intensive and imprecise, creating a bottleneck in selecting desirable traits. This studys main objective is to investigate soybean resistance to stink bugs through innovative strategies based on high-throughput phenotyping (HTP) using unmanned aerial vehicles (UAVs), genome-wide association studies (GWAS), and genomic selection (GS). We used 290 soybean lines, phenotyped over two to three growing seasons under natural stink bug infestations. UAVs equipped with RGB cameras captured aerial images at different flight periods. The manually measured phenotypic traits were correlated with color, texture, and color histogram indices derived from the images. Machine learning (ML) modelsAdaBoost, SVM, and MLPwere tested to predict these traits. GWAS was performed to identify SNPs associated with the phenotypic traits, while multi-trait GS models were developed to predict resistance and productivity. Vegetative indices, especially the VARI index, and texture-based indices showed high correlation with manually measured traits in highly stressful environments. However, ML models were less effective in predicting stink bug resistance under high infestation levels. GWAS identified 71 significant SNPs, with genes annotated in 52 of these regions, distributed across almost all soybean chromosomes. Specific SNPs showed a high explanation of phenotypic variation, such as GM02_19807216 and GM12_3274686, and many were associated with both types of traits. The results highlight specific genomic regions and candidate genes that can be targeted in future programs, promoting significant advances in soybean resistance to stink bugs. The integration of image-based phenotyping with GS significantly increased the predictive ability of the models, especially in environments with high pest pressure. This study is pioneering in the field and demonstrates the feasibility and effectiveness of using image-based traits for phenotyping stink bug resistance in soybean breeding programs. The integration of HTP in genomic studies offers an effective approach to accelerate the selection process, reducing costs and time, and allowing for more efficient and precise breeding strategies. The results provide a solid foundation for future research and development of more resistant and productive soybean cultivars.