Seleção genômica e fenômica para identificação e seleção de progênies superiores de soja

Abstract

Genomic selection has been supporting breeding programs to increase genetic gains. However, innovative strategies such as phenomic selection, which employs Near- Infrared Spectroscopy (NIR) to record reflectance values of the chemical composition of samples for predicting complex traits, have proven advantageous when integrated with genomic prediction models. In order to investigate prediction models considering molecular markers and NIR spectra, this study aimed to assess and compare different genomic, phenomic, and combined prediction models to estimate the genetic potential of soybean progenies, targeting the advancement of inbreeding generations. The data were obtained from the assessment of F2:3 soybean progenies belonging to a population developed at the Federal University of Lavras (UFLA). The experiment was conducted at two locations in the state of Minas Gerais, Lavras and Ijaci, using a square lattice design. The traits assessed included days to full maturity (DFM), total number of pods (TNP), total number of grains (TNG), and grain yield (Yield). Phenotypic data were analyzed using mixed models. All plots were genotyped for marker acquisition and estimation of Genomic Estimated Breeding Values (GEBVs). NIR spectra were obtained from bulked seed samples collected from both locations. After quality control, a total of 3149 SNPs and 1500 wavelengths were retained. Predictive capacity and coincidence index (CI) were estimated based on the predicted genetic values derived from the different strategies. Individual and multi trait selection were performed using the Mulamba and Mock index. The results demonstrated the potential of these strategies for both selection and discarding of progenies. Among the different prediction models studied, it can be inferred that the highest predictive capacities were achieved through the genomic model and the models integrating both marker and spectral data, indicating that the incorporation of phenomic data obtained by spectroscopy enhances the predictive capacity of the models.