Enviromic-based kernels optimize resource allocation with multi-trait multi-environment genomic prediction for tropical maize

• Autor(a) principal: Gevartosky, Raysa
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• 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-07012022-094055/
• topic: Capacidade preditiva; Caracterização ambiental; Envirotyping; Genomic selection; Optimized training set; População de treinamento otimizada; Predictive ability; Response to selection; Resposta à seleção; Seleção genômica

Genomic prediction (GP) success is directly dependent on establishing a training population. Incorporating high-quality envirotyping data increases the efficiency of GP models, especially for multi-environment trials, and provides a better explanation of variation sources. Thus, it can help on multi-trait multi-environment trials (MTMET) by improving predictive ability (PA), selecting information more assertively, and capturing relationships between environments and genotypes. Therefore, in this study, we aimed to design optimized training sets for MTMET. The phenotypic labor is diminished due to lower but optimally selected population sizes while keeping the predictive ability at satisfactory levels. For that, we evaluated the predictive ability of five GP models using the Genomic Best linear unbiased predictor model (GBLUP) with additive + dominance effects (M1) as the gold standard and then adding genotype by environment interaction (G × E) (M2), enviromic data (W) (M3), W+G × E (M4), and finally W+G × W (M5), where G × W denotes the genotype by enviromic interaction. Moreover, we considered single-trait multi-environment trials (STMET) and MTMET, for three traits: grain yield (GY), plant height (PH), and ear height (EH), with two datasets and two cross-validation schemes. Afterward, we built two kernels for genotype by environment by trait interaction (GET) and genotype by enviromic by trait interaction (GWT) to apply genetic algorithms to select genotype:environment:trait combinations that represent 98% of the variation of the whole dataset and composed the optimized training set (OTS). Then, we performed GP and accessed its PA and genetic gain per amount invested. Subsequently, we compared benchmarks with OTS regarding the PA and genetic improvement per unit invested. Considering the best scenario for OTS, which included the GWT kernel, there was a reduction of up to 60% in terms of PA. On the other hand, it was possible to reduce the number of plot:traits to be phenotyped up to 98%. Furthermore, using OTS based on enviromic data, it was possible to increase the response to selection per amount invested by 142%. Consequently, our results suggested that genetic algorithms of optimization associated with genomic and enviromic data are efficient in designing optimized training sets for genomic prediction and improve the genetic gains per dollar invested. Although, it is worth remembering that exist specific interactions within datasets that should not be ignored when using the proposed approach.