PROBABILISTIC IDENTIFICATION OF CORYNEFORM BACTERIA

A frequency matrix of positive results was constructed for major phena defined in a previous phenetic classification of several species of c oryneform and related taxa. A total of 280 physiological characters fr om 441 strains were taken as basis for this identification matrix. Min imum number of diagnostic characters for the matrix was selected using computer programs for calculation of different separation indices (CH ARSEP) and selection of group diagnostic properties (DIACHAR). The res ulting matrix consisted of 31 phena versus 58 characters. Phena overla p within the matrix was found to be relatively small (OVERMAT program) . For each phenon, the identification scores for the most typical hypo thetical organism was satisfactory (MOSTTYP program). The matrix was e valuated theoretically and practically (MATIDEN program). The overall identification rate (442 strains) of the theoretical evaluation (Willc ox probability >0.99) was 92.0%. In the practical evaluation (40 strai ns) a total of 33 strains (82.5%) were identified with a Willcox proba bility >0.9. For minor clusters and single strains belonging to geneti cally and chemotaxonomically defined species, additional identificatio n tables are provided. When used in combination with chemotaxonomic me thods, the identification matrix can improve the identification of cor yneform bacteria.