This study aimed to identify the genetic divergence among coffee trees of the Maragogipe group in the active germplasm coffee bank in Patrocínio, Minas Gerais, Brazil, using multivariate analyzes. In the study, 27 morpho-agronomical descriptors were assessed. Mahalanobis generalized distance was applied to quantify the genetic divergence among genotypes. The following methods were used: Tocher clustering, Unweighted Pair-Group Method Using Arithmetic Averages (UPGMA) hierarchical clustering and the canonical variables analysis. The cluster analysis by Tocher method and the UPGMA separated the accessions into three and six groups, respectively. The analysis of the relative contribution of each characteristic to the genetic dissimilarity highlighted the intensity of curling leaf, inflorescence number per leaf axil and average yield as those that contributed most to the achievement of genetic divergence. It was observed that the first three canonical variables explained more than 90% of the total variance contained in the set of the analyzed characteristics. The combination between MG0167 and MG0170 accessions is the most promising for a possible crossing for immediate use in breeding programs.
Coffea arabica; multivariate analysis; genetic breeding