A probability matrix for identification of some Streptomycetes

Area of Science:

• Microbiology
• Numerical Taxonomy
• Bacterial Identification

Background:

• Phenetic numerical classification of bacteria, particularly Streptomyces, relies on character state data.
• Previous classification by Williams et al. (1983) defined clusters at 77.5% SSM similarity.
• A need exists for objective methods to identify and group the numerous described Streptomyces species.

Purpose of the Study:

• To construct a probabilistic identification matrix for bacterial clusters.
• To evaluate the theoretical and practical soundness of the identification matrix.
• To assess the matrix's utility in identifying unknown bacterial isolates.

Main Methods:

• Utilized character state data from 23 phena (19 Streptomyces, 2 Streptoverticillium, 'Nocardia' mediterranea, Streptomyces fradiae).
• Selected diagnostic characters using Sneath's CHARSEP and DIACHAR programs, creating a 41x23 character-phena matrix.
• Evaluated matrix performance using Sneath's MATIDEN program, including cluster overlap (OVERMAT) and Hypothetical Medium Organism scores (MOSTTYP).

Main Results:

• The identification matrix was theoretically sound, with acceptable test error.
• Practical evaluation confirmed matrix reliability and minimal distortion of identifications.
• Successfully identified 80% of unknown isolates from various habitats with a specific cluster.

Conclusions:

• The developed probabilistic identification matrix is a reliable tool for bacterial classification.
• The matrix demonstrates theoretical soundness and practical efficacy in identifying unknown isolates.
• This matrix can serve as a foundation for more objective identification and grouping of Streptomyces species.