The common bean breeding is based, mainly on the hybridization of cultivars and lines, so it is important to know the dissimilarity among the parents to obtain superior genotypes. The objectives of this work were: to cluster the genotypes by multivariate techniques using hierarchical method (nearest neighbor) and the canonical variables; to point the relative contribution of the analyzed traits for genetic dissimilarity and to point out the most promising hybrid combinations to produce superior recombination using as dissimilarity measure the square of the Euclidean distance. Thus, thirteen bean genotypes (Phaseolus vulgaris L.), involving commercial varieties and advanced lines, were assessed at the experimental Campus at the Universidade Estadual de Mato Grosso do Sul - UEMS, in Aquidauana, Mato Grosso do Sul State, Brazil. The experiment was carried out in a randomized complete block design with three replications in the "dry" crop season, in the period 2005/2006. There were assessed the following traits: number of pods per plant (POD), number of seeds per pod (SEED), mass of 100 seeds (MHS) and grain yield (YD). The beans genotypes were clustered in three groups by nearest neighbor method and in five groups by canonical variables. The trait "MHS" had the highest relative contribution to total genetic dissimilarity. The cultivar IAC Carioca Eté was the most dissimilar and it can strategically be used in directed artificial hybridizations. The most divergent association were "CNFv 8025 x IAC Carioca-Eté", "Rudá x IAC Carioca-Eté" and "CNF 7135 Bambuí x IAC Carioca-Eté".
Phaseolus vulgaris L.; multivariate analyses; genetic variability
