IDENTIFYING SUBSETS OF MAIZE ACCESSIONS BY 3-MODE PRINCIPAL COMPONENTANALYSIS

Genebank accessions are a potential source of genetic variability for maize (Zea mays L.) breeding programs. Identification of useful indivi dual entries is commonly based on the expression of one or more attrib utes in different sites or environments. This study was motivated by t he need to identify a subset of Caribbean accessions for introgression into elite germplasm of the western Corn Belt. Thus the objective,vas to identify potentially useful Caribbean accessions, based on (i) sim ultaneous consideration of six agronomic attributes deemed economicall y important for the western Corn Belt, and (ii) response patterns obse rved across four sharply contrasting environments. Both (i) and (ii) w ere addressed by means of three-mode principal component analysis (PCA ) of data on six agronomic and morphological attributes for 184 Caribb ean maize accessions evaluated at four environments. Three-way data we re analyzed by three-mode PCA, and based on (i) and (ii), two joint pl ots were generated. From the PCA and joint plots, two subsets of acces sions were identified. First, a subset of 14 accessions having good yi eld, intermediate plant height, and average days to anthesis was ident ified. Secondly, a subset of 10 accessions having average performance over all environments was identified. Two accessions were common to bo th subsets. Jointly, the two approaches produced a combined subset of 22 accessions, representing 12% of the total evaluated, and including representatives of 11 maize races. Three-mode PCA integrated informati on on accessions, attributes, and environments, and provided a means o f simultaneously visualizing these three types of information. Three-m ode PCA can complement standard methodologies used by plant breeders f or identification of potentially useful accessions in introgression pr ograms.