Soil bacteria: principal component analysis of descriptions of named cultures.

Abstract

Binary-encoded descriptions of 85 named cultures (mostly of soil origin) and of 15 unnamed soil isolates were analyzed by a two-stage principal component procedure including the condensation of the data in terms of grouping of tests and of cultures. This procedure also allowed an evaluation of the discriminative importance of the tests used. Nine clusters of cultures were found by use of a centroid clustering procedure in the space defined by the five leading principal component (P.C.) vectors, which collectively accounted for 54% of original variance. The clusters of item-points representing the cultures were plotted in the dimensions defined by the first and second P.C. vectors providing a geometrical model by which their relationships could be visualized. These P.C. vectors could be interpreted in terms of 31 important discriminative tests (many of which were concerned with the utilization of carbon compounds) from the total of 98 tests used. Fifteen clusters of tests were found as a result of Adansonian R analysis, and 37 clusters of cultures found by means of Adansonian Q analysis. The groupings of the isolates by the Q analysis and the P.C. analysis were compatible although there was some geometrical distortion. To supplement personal decisions on the acceptability of automatically derived clusterings, additional analyses were performed to investigate the homogeneity, relative sizes, and spatial relationships of the clusters of organisms found. In general, results of taxonomic interest were concordant with those of other workers, especially in respect of the genera Arthrobacter, Rhizobium, and Pseudomonas. It was concluded that the bacteriological and the numerical methods used would be appropriate for comparative study of heterotrophic bacterial populations from rhizosphere and corresponding soil environments.