ABSTRACT. The aim of this study was to evaluate the phenotypic variation and diversity of cassava for breeding purposes based on the pasting properties of starch, for food applications. The viscosities of the starches, extracted from 1031 accessions, were obtained using a Rapid Visco Analyzer. The best linear unbiased predictors were calculated for five critical points of the viscograms, which were then used to cluster the accessions based on the discriminant analysis of principal components. The wide phenotypic variation identified from the viscograms reveals the high potential for selection gains, especially for pasting temperature and setback. Certain strong correlations indicate that selection based on a specific viscogram trait can be used for indirect selection (e.g., the correlation between peak viscosity and breakdown [0.85]). The lowest Bayesian information criterion identified five different groups. Groups 3 and 4 exhibited high pasting temperatures, while Groups 3 and 5 exhibited low setbacks. Group 3 also exhibited low viscosity and breakdown. However, we focused on selecting cassava accessions with a high pasting temperature and low setback, as these are the most important traits for industrial applications. The predicted genetic gains from the selection of the top 15 cassava accessions for increasing pasting temperature and decreasing setback were 4.40% and 53.17%, respectively. The increased pasting temperature (~72.89°C) and high amplitude of setback (~600 cP) due to selection can guide the utilization of the cassava germplasm for breeding and provide a basis for further studies to develop varieties with added value, especially in the food industry.

ABSTRACT: Genetic redundancy in cassava (Manihot esculenta Crantz) presents a challenge to efficient management of genetic resources. This study aimed to identify and define the genetic structure of duplicates in cassava germplasm from various Embrapa research units, using single-nucleotide polymorphism (SNP) markers. We evaluated 2,371 accessions with 20,712 SNPs. The identification of duplicates was performed based on multilocus genotypes (MLG), adopting a maximum genetic distance threshold of 0.05. The population structure was defined based on discriminant analysis of principal components (DAPC). A total of 1,757 unique and 614 duplicate accessions were identified. The redundancy of the collections ranged from 17 % (Belém, PA – Brazil) to 39 % (Petrolina, PE – Brazil), with an average of 21 %. This redundancy between different research units is probably due to the historical sharing of accessions, as well as collections carried out in the same region, or even to the intense germplasm exchange between farmers with different genotype names. In terms of genetic structure, the 250 principal components explained 88 % of the genetic variation of the SNP markers and defined the hierarchical structure of the duplicate cassava germplasm in 12 groups. Since heterotic groups have not yet been identified for cassava, crosses between accessions of the 12 DAPC groups may be promising. All MLGs were allocated within the same DAPC group, corroborating duplicate analyses yet still revealing high variability between groups that were quite distinct based on the first two discriminant functions. Our results contribute to optimizing the conservation of genetic resources, together with understanding diversity and its use in crop improvement.