Development of cassava core collections based on morphological and agronomic traits and SNPS markers

Cassava (Manihot esculenta Crantz) holds significant importance as one of the world's key starchy crop species. This study aimed to develop core collections by utilizing both phenotypic data (15 quantitative and 33 qualitative descriptors) and genotypic data (20,023 single-nucleotide polymorphisms) obtained from 1,486 cassava accessions. Six core collections were derived through two optimization strategies based on genetic distances: Average accession-to-nearest-entry and Average entry-to-nearest-entry, along with combinations of phenotypic and genotypic data. The quality of the core collections was evaluated by assessing genetic parameters such as genetic diversity Shannon-Weaver Index, inbreeding (Fis), observed (Ho), and expected (Hs) heterozygosity. While the selection of accessions varied among the six core collections, a seventh collection (consolidated collection) was developed, comprising accessions selected by at least two core collections. Most collections exhibited genetic parameters similar to the complete collection, except for those developed by the Average accession-to-nearest-entry algorithm. However, the variations in the maximum and minimum values of Ho, Hs, and Fis parameters closely resembled the complete collection. The consolidated collection and the collection constructed using genotypic data and the Average entry-to-nearest-entry algorithm (GenEN) retained the highest number of alleles (>97%). Although the differences were not statistically significant (above 5%), the consolidated collection demonstrated a distribution profile and mean trait values most similar to the complete collection, with a few exceptions. The Shannon-Weaver Index of qualitative traits exhibited variations exceeding ±10% when compared to the complete collection. Principal component analysis revealed that the consolidated collection selected cassava accessions with a more uniform dispersion in all four quadrants compared to the other core collections. These findings highlight the development of optimized and valuable core collections for efficient breeding programs and genomic association studies.