An approach to numerical identification of bacterial species

The distribution of matching coefficients (M values) for strains of a species to their own 'hypothetical mean organism' (HMO) or to HMO patterns of other organisms was studied in 754 strains of 19 mycobacterial species, testing for 91 discriminating characters. The M values of strains of a species to the HMO of other species usually showed a normal distribution, and M values to their own HMO showed either a normal distribution of a binomial distribution, depending on the mean of M values. If the number of test characters was large, the binomial distribution usually resembled the normal distribution. After preparation of the HMO for every species and estimation of the mean of the M values (M) and the standard deviation (s), numerical identification could be carried out: if a test strain had an M value to the HMO of species chi that only fell within the range (M +/2s) for species chi, the strain would be identified as a member of that species.