Identification of Bacteria by Computer: General Aspects and Perspectives

SUMMARY: The general problems of probabilistic identification of bacteria are discussed, particularly choice of taxa, choice of tests, assignation of probabilities and variation with geographical distribution, method of modification of probabilities as information is obtained, identification levels and linkage of tests. A study is described of the probabilistic identification of 1079 reference strains, and 516 field strains of Gram-negative, aerobic, rod-shaped bacteria. The field strains were identified both on the test results obtained by sending laboratories and test results in our laboratory. Identification rates for fermentative bacteria of 90.8% for reference strains and 89.4% for field strains were obtained, and for non-fermentative bacteria of 82.1% and 70.8% respectively. The field strains were received because they were difficult to identify in the medical diagnostic laboratory; higher rates of identification might be expected for typical strains.