A probability matrix for identification of some Streptomycetes

The character state data obtained for clusters defined at the 77.5% SSM similarity level in the phenetic numerical classification described by Williams et al. (1983) were used to construct a probabilistic identification matrix. The 23 phena included were the major clusters (19 Streptomyces, 2 Streptoverticillium and 'Nocardia' mediterranea) and one minor cluster (Streptomyces fradiae). The characters most diagnostic for these clusters were selected using Sneath's CHARSEP and DIACHAR programs. The resulting matrix consisted of 41 characters x 23 phena. Identification scores, determined by Sneath's MATIDEN program were used to evaluate the matrix. Theoretical assessment was achieved by determination of the cluster overlap (OVERMAT), the identification scores for the Hypothetical Medium Organism of each cluster (MOSTTYP), and the scores for randomly selected cluster representatives using the classification data of Williams et al. (1983). The matrix was evaluated practically by the independent re-determination of the characters for the same cluster representatives, which also provided a measure of test error. Finally it was used to identify unknown isolates from a range of habitats. The results showed that the matrix was theoretically sound. Test error was within acceptable limits and did not distort identifications. Of the unknown isolates, 80% were clearly identified with a cluster. It is suggested that the matrix could form the basis for a more objective identification and grouping of the large number of Streptomyces species which have been described.