Agronomic, biochemical and molecular diversity among cassava accessions (Manihot esculenta Crantz) collected in different regions in BrazilA relevant species as food source for the world population, especially for the underdeveloped and emerging countries, is cassava (Manihot esculenta Crantz). Primarily, cassava is a provider of energy from starch accumulated in its roots, but the presence of carotenoids with pro-vitamin A activity is also important. In this context, the objective of this study was to characterize the agronomic, biochemical and molecular diversity of cassava accessions selected for the color of their roots, collected in the Amazon, Central-West, Southeast and South regions in Brazil. The characterization was performed from 81 genotypes (33 from the Amazon region, 24 from Central-West, 18 from Southeast, four from the South and two commercial varieties) using 16 agronomic traits, five biochemical characters and 17 microsatellite loci. For biochemical and agronomic traits analyses of variance, Tukey test and Mahalanobis distances were obtained considering the region of origin of the genotypes. For microsatellite markers, indices of genetic diversity were estimated, such as number of alleles per locus (A), percentage of polymorphic loci (P), observed (H o ) and expected (H e ) heterozygosities, as well as the structuring of diversity between and within groups (H S , H T , D ST and G ST ). Also, a cluster analysis was obtained using the Neighbor-Joining method and Nei´s genetic distance, as well as a Bayesian analysis. Moreover, the distances obtained for all characters were correlated using the Mantel test. Results indicate high variability for all characters evaluated. For most of agronomic traits there was significant variation between and within regions. All biochemical characters showed highly significant variation between and within regions. Taking into account the agronomic characters, there was a tendency for the Amazon genotypes to be grouped together, where the characteristics total plant height, height of the first branch, relationship between these characteristics and harvest index were those that mostly determined the genotypes distribution. For the biochemical characters, the Amazon genotypes were also grouped due to the high concentration of cyanogenic compounds and total carotenoids. Molecular markers indicated high genetic variability (A = 3.58; P = 100%; H o = 0.535; H e = 0.642), while most of total diversity (H T = 0.651) was found within regions (H S = 0.643). In the cluster analysis a tendency was observed towards a structured genetic variability, with the grouping of most genotypes collected in the Amazon region. The Bayesian analysis separated the genotypes in two groups, with one of the groups including most of the Amazon genotypes, in accordance with the cluster analysis. The correlations between the distances obtained for all characters were not significant. Considering the genotypes content of cyanogenic compounds and the distance between them from the microsate...