A CLASSIFICATION OF 148 US MAIZE INBREDS .2.. VALIDATION OF CLUSTER-ANALYSIS BASED ON RFLPS

To be useful to breeders, classification of genotypes based on cluster analysis must provide meaningful groupings of the genotypes clustered . We evaluated a classification of 148 U.S. maize [Zea mays L.] inbred s resulting from cluster analysis based on restriction fragment length polymorphisms (RFLPs) to determine if it represented the true associa tions among the lines. Testing was aimed at the products of the two st eps in cluster analysis: the proximity matrix containing estimates of relationship computed from the data and the phenogram displaying group s in the form of a tree diagram. The proximity matrix and a matrix of pedigree relationships were compared by the Hubert Gamma statistic. Di ssimilarities indicated in the phenogram were correlated with those de fined in the proximity matrix. The grouping displayed in the phenogram was compared to that exhibited in phenograms resulting from three add itional cluster analyses generated by different methods for computing proximities. These groupings were then compared to the expected groupi ng based on pedigree information. The patterns present in the proximit y matrix were substantiated by pedigree information. Based on agreemen t between the phenogram and the proximity matrix, the phenogram depict ed estimates of genetic relationship accurately. Inbreds were grouped similarly in the four classifications and the level of correspondence of inbred group assignments to the expected grouping based on availabl e pedigree information was similar across classifications, suggesting that a natural grouping of the Lines exists and was generally reflecte d in each classification. Therefore, the classification was judged to reasonably represent the true associations among the 148 maize inbreds . In addition, the advantages of a method to compute proximities by a formula proposed by Nei and Li were noted.