APPLICATION OF MINIATURIZED PHYSIOLOGICAL TESTS IN NUMERICAL CLASSIFICATION AND IDENTIFICATION OF SOME BACILLI

Four hundred and sixteen strains of aerobic, endospore-forming bacteria belonging to the genus Bacillus were physiologically characterized using a total of 329 miniaturized tests. Overall similarities of all strains for 295 unit characters were determined by numerical taxonomic techniques using the UPGMA algorithm and the S(SM) and the S(J) coefficients as measures of similarity. The test error and cophenetic correlation were within acceptable limits. Comparison of photometric and visual test reading revealed overall differences of 12.3%. A total of 30 clusters (2 or more strains) and 20 single member clusters were defined at the 86.7 to 89.40% levels (S(SM)). Generally similar groupings were obtained with the Jaccard coefficient with some changes in the definition of clusters. Heterogeneity was found within the species Bacillus brevis, B. circulans, B. macerans, B. coagulans, B. megaterium, B. sphaericus and B. stearothermophilus, thus confirming the results of previously published data. For the clusters and for some subclusters containing two or more strains the minimum number of differentiating characters was selected by computer programs (CHARSEP, DIACHAR, MOSTTYP) and a frequency matrix for probabilistic identification was constructed consisting of 40 tests versus 36 clusters and subclusters. This matrix was theoretically evaluated using a computer program (MATIDEN), and out of 391 strains, a total of 287 (73.40%) were correctly identified with a Willcox probability p > 0.99. Another 34 strains (8.69%) achieved Willcox probabilities above 0.90, for 19 strains (4.85%) p exceeded 0.8, and 27 strains (6.90%) achieved Willcox probabilities > 0.5. For 24 strains (6.130%) no correct identification result was obtained.