Research Article

Colored fiber cotton in the Uberlândia region using artificial neural networks for yield assessment

Published: February 25, 2019
Genet. Mol. Res. 18(1): GMR18104 DOI:
Cite this Article:
D.B.O. Cardoso, E.G.Silva Júnior, M.C. Reis, L.B. de Sousa, L.M.S. Falco, M.C. Mamede, B.Q.V. Machado, T.S. Paiva, C.P. Gundim, G.A. Alves (2019). Colored fiber cotton in the Uberlândia region using artificial neural networks for yield assessment. Genet. Mol. Res. 18(1): GMR18104.


Cotton is the most widely utilized natural fiber in the world. Brazil is currently one of the world’s largest cotton producers. Cotton crops are cultivated in all regions of the country, especially in the Cerrado biome. Studies of genotype x environment (GxE) interactions evaluate the adaptability and stability of cotton genotypes. Adaptability and stability evaluations help understand genotype responses to environmental stimuli and the predictability of genotypes in their response to environmental oscillations. We examined the effect of the genotype x environment interaction on cotton yield and fiber characteristics and compared artificial neural networks (ANNs) with conventional methods for assessing adaptability and stability of colored-fiber cotton genotypes. The experiment was conducted at the experimental farm of Universidade Federal de Uberlândia, during four crop years. Twelve genotypes of colored-fiber cotton were evaluated. The experimental design was randomized complete blocks with three replicates. Seed cotton yield was evaluated. The GxE interaction was analyzed by the F-test at α = 0.05. Adaptability, stability, and the factors of the decomposed GxE interaction were analyzed by the Eberhart and Russell, Centroid and ANN methods. The GxE interaction was significant for the variable seed cotton yield, demonstrating differences in genotype behavior among environments. The interactions were predominantly complex. There was concordance between Eberhart and Russsell and ANN analyses. Genotypes UFUJP-02 and UFUJP-17 were responsive to environmental stimuli; they had high predictability, in addition to high fiber yield. The ANN method reliably evaluated adaptability compared with Eberhartand Russel and Centroid methods.