Factor analysis applied in genomic prediction considering different density marker panels in rice

The study objective was to evaluate the application of factor analysis (FA) on genomic prediction considering different density marker panels. The FA transforms phenotype traits in latent variables (factor scores), called pseudo-phenotype in this study. The Genomic Best Linear Unbiased Prediction method was applied to the Oriza sativa L phenotype traits. The dataset contains twenty-two phenotypic traits and 36,901 SNPs (Single Nucleotide Polymorphism) from 413 genotypes. The results obtained indicate that combining the factor analysis and the genomic prediction with different density marker panels was efficient. The analysis presented similar values for predictive ability, considering the phenotypes and pseudo-phenotypes (in both analyses, there was variation between 0.60 and 0.80), high agreement of SNPs with major effects, and high agreement between the best and worst selected individuals considering phenotypes and pseudo-phenotypes analysis.