Assessment of genetic diversity in three subsets constituted from the ICRISAT sorghum collection using random vs non-random sampling procedures A. Using morpho-agronomical and passport data

Abstract

A large collection, such as the sorghum [Sorghum bicolor (L.) Moench] landrace collection held at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), represents a challenge for the maintenance of both the accessions of and the information documented for the germplasm collection. The accessibility and knowledge of the landrace collection are the essential factors for an efficient utilization of the genetic resources by both breeders and farmers. Different sampling strategies, either random or non-random, were proposed to obtain subsets of reduced size (core collection). Three subsets were established; a random sampling within a stratified collection (logarithmic strategy: L); a sample based upon morpho-agronomic diversity (principal component score strategy: PCS); and a sample based upon an empirical knowledge of sorghum (taxonomic strategy: T). Comparisons of these three samples for morpho-agronomic characterization and passport information were assessed to determine their impact on phenotypic diversity. For their overall diversity, the three subsets did not differ, as shown with the two-dimensional representation of the morpho-agronomic diversity and the Shannon-Weaver diversity indices. When comparisons for morpho-agronomic and passport data were considered, the PCS subset looked similar to the entire landrace collection. The L subset showed differences for characters associated with the photoperiod reaction that was considered in the stratification of the collection. The T subset was the most distinct from the entire landrace collection as it over-represented the landraces selected by farmers for specific uses and covered the widest range of geographical adaptation and morpho-agronomic characteristics.