Ferreira de Oliveira, Gustavo HugoMurray, Seth C.Cunha Júnior, Luis CarlosGomes de Lima, Kássio Michellde Lelis Medeiros de Morais, CamiloHenrique de Almeida Teixeira, Gustavo [UNESP]Môro, Gustavo Vitti [UNESP]2020-12-122020-12-122020-01-01Journal of Cereal Science, v. 91.1095-99630733-5210http://hdl.handle.net/11449/201282One of the most important quality traits in popcorn breeding programs is the popping expansion (PE) capacity of the kernel, which is the ratio of the volume of the popcorn to the weight of the kernel. In this study, we evaluated whether near infrared spectroscopy (NIR spectroscopy) could be used as a tool in popcorn breeding programs to routinely predict and/or discriminate popcorn genotypes on the basis of their PE. Three generations (F1, F2, and F2:3) were developed in three planting seasons by manual cross-pollination and self-pollination. A total of 376 ears from the F2:3 generation were selected, shelled, and subjected to phenotypic analysis. Genetic variability was observed in the F2 and F2:3 generations, and their average PE value was 31.5 ± 6.7 mL g−1. PE prediction models using partial least square (PLS) regression were developed, and the root mean square error of calibration (RMSEC) was 6.08 mL g−1, while the coefficient of determination (RC 2) was 0.26. The model developed by principal component analysis with quadratic discriminant analysis (PCA-QDA) was the best for discriminating the kernels with low PE (≤30 mL g−1) from those with high PE (>30 mL g−1) with an accuracy of 78%, sensitivity of 81.2%, and specificity of 72.2%. Although NIR spectroscopy appears to be a promising non-destructive method for assessing the PE of intact popcorn kernels for narrow breeding populations, greater variability and larger sample sizes would help improve the robustness of the predictive and classificatory models.engDiscriminationMultivariate analysisPredictionSelection methodsZea mays L.Artigo10.1016/j.jcs.2019.1028612-s2.0-85074165239