A review of numerical methods in bacterial identification

Miniaturized tests for computer-assisted identification of motile Aeromonas species with an improved probability matrix

“…Development of the probability matrix was undertaken using the data set developed by Oakey et al (1996) and modi®ed by incorporating data for new tests and species from published reports (Willcox et al 1980). The ®rst revision of the probability matrix was tested using the type and de®nition strains of the 12 motile genospecies, but in these initial assessments, the Willcox probability score, P, ranged from 0á85 to 0á95.…”

“…Inspection of the data set revealed several anomalies between test and published data, as well as reactions of some type strains observed in the miniaturized tests. Accordingly, data sets were modi®ed to accommodate these changes based on test performance and strain variation (Willcox et al , 1980. Characterization of Aeromonas species has been undertaken largely at 30 or 37°C, yet there is evidence that strains may react differently at 37, 30 or 25°C (Altwegg et al 1987;Wakabongo et al 1992;Ha Ènninen 1994;Ali et al 1996), possibly with profound effects on identi®cation outcome.…”

New Strategies for the Selective Isolation of Industrially Important Bacteria

“…The theoretical basis of numerical identification and the use of this method to identify streptomycetes have been reviewed in detail elsewhere (Hill, 1974;Willcox, Lapage and Holmes, 1980;Williams, Vickers and Goodfellow, 1985). The streptomycete probabilistic matrix was based on 23 clusters which included all of the major clusters and Streptomyces fradiae, a well-known source of antibiotics.…”

The characterisation of Bordetella/Alcaligenes‐like organisms and their effects on Turkey Poults and chicks

“…Kersters et al (1984) concluded that Alcaligene faecalis, Bordetella bronchiseptica and B. avium, (isolate B) were three distinct taxa; their researches included similar biochemical tests, but were influenced particularly by electrophoresis of soluble bacterial proteins and nucleic acid analyses, neither of which were undertaken in the current study. The outcome of numerical analysis depends on the choice and number of isolates and tests, on test reproducibility, and in addition, on the choice of computer programme (Jones and Sackin, 1980), and since it also relies on the independence of each character, Willcox et al (1980) suggests that closely related tests should be excluded. In view of these comments, the API data is particularly difficult to assess, for amongst the many substrates screened, it is likely that some may be linked through the same biochemical pathway, or through a particular enzyme, or maybe related chemically thereby biasing the outcome of the analysis.…”