Research Article

Analysis of genetic divergence in sweet corn genotypes through hierarchical optimization methods

Published: August 22, 2019
Genet. Mol. Res. 18(3): GMR18384 DOI:
Cite this Article:
F.J. Carvalho, G.M. Maciel, O.J. Marques, I.G. da Silva, G.D. Braga, G.R. Marquez, A.C.S. Siquieroli, L.D.C. Neto (2019). Analysis of genetic divergence in sweet corn genotypes through hierarchical optimization methods. Genet. Mol. Res. 18(3): GMR18384.


Sweet corn (Zea mays subsp. saccharata) is considered a special vegetable of high nutritional value. One of the barriers encountered by breeders has been a lack of adequate genetic variability of sweet corn, coupled with a need for appropriate methodologies to evaluate the existing genetic diversity. Our objective was to determine the best method to identify promising genotypes to improve sweet corn production. We used data from 181 open-pollinated sweet corn genotypes, cultivated during 2016 and 2017. Multivariate analyses were carried out to determine the genetic dissimilarity between the genotypes, obtaining the matrix of dissimilarity by Euclidean distance. Prior to calculating the distance between matrices, two data standardizations (Z1 and Z2) were performed for comparison. Genetic divergence was analyzed by four distinct hierarchical methods: Unweighted Pair-Group Method Using Arithmetic Averages (UPGMA), Ward, Weighted Pair-Group Method Using Arithmetic Averages (WPGMA) and Single Linkage. Tocher’s optimization method was also used. The Simple Linkage and UPGMA methods presented similar groupings, consistent with breeding program aims and with the highest values of co-phenetic correlation coefficient (CCC). The Ward’s method was not efficient, because it produced several clusters without isolating different genotypes. Furthermore, it was the method with the lowest CCC for both matrices. The standardized Z2 matrix should be avoided, especially when a large number of genetic traits are measured, in order to prevent possible overlapping between traits, a variables with higher standard variations could contribute more to the clustering.