Comparative Assessment of Variation among Sorghum Germplasm Accessions Using Seed Morphology and RAPD Measurements

The sorghum germplasm collection currently contains over 42 000 accessions, a number that is too large to manage efficiently. The specific objective of this research was to compare clusters developed from agronomic descriptors with phylogenetic groupings based on random amplified polymorphic DNA (RAPD) fingerprinting of selected sorghum [Sorghum bicolor (L.) Moench] races. Our intent was to identify one approach using agronomic descriptors that would most closely approximate the groupings produced by RAPD markers. Ninety-four accessions of sorghum were grouped into four of the five major races. Differences among accessions determined by various clustering procedures based on agronomic characteristics were compared with clusters developed by means of RAPD markers. Each race varied in the degree of similarity between the four clustering approaches taken and the information provided by RAPD fingerprinting. Test 2, standardization of data by Z-scores and cluster analysis using the complete set of data, provided the highest similarity score for the race bicolor, while Test 3, standardization of data by Z-scores and cluster analysis based on a reduced set of variables selected from principle component analysis, provided the highest similarity scores for the races guinea. Test 1, random selection, was highest for the races caudatum and durra. When averaged over all the races, Test 2 provided the highest similarity score. The results of this study indicate that no one approach to develop clusters by means of agronomic descriptors closely approximate the groupings produced by RAPD markers. These results underscore the need for further research in the evaluation of techniques used to develop core collections and their validity.