engleski

Identification of enterococcal strains using an inductive learning algorithm

A total of 153 enterococcal isolates from blood and urine were collected at the University Hospital for Infectious Diseases, Zagreb, Croatia in the period from March 1992 to September 1992. One hundred and thirty-eight of the isolates were identified as E.faecalis (90.1%), 13 (8.4%) as E. faecium and the remaining 2 (1.5%) as E. avium by the conventional test scheme (Facklam, Collins). The aim of this study was to evaluate a simple scheme for species identification of most frequently isolated enterococcal strains in our institution. The necessary and sufficient conditions to detect E. faecalis, E. faecium and E. avium are computed from the selected set of 13 tests by the inductive learning algorithm ILLM (Inductive Learning by Logic Minimization). The tests were: 1) growth in 6.5% NaCl, 2) tolerance to 0.4% tellurite, 3) reduction of 0.01% tetrazolium, 4) detection of yellow pigment, 5) arginine hydrolysis, 6) hydrolysis of sodium hippurate, 7) fermentation of sucrose, 8) lactose, 9) raffinose, 10) glycerol, 11) sorbitol, 12) mannitol and 13) determination of Lancefield's D antigen. The ILLM algorythm searches over all possible combinations of the tests from starting set and selects the minimal condition with the property that is true for and only for the target species. We analysed 16 enterococcal species and found out 18 tests that correctly identified E. faecalis, E.faecium and E. avium. The results were: E. faecalis( if (50=neg, 9=neg, 10=pos) ; E.faecium if (2=pos, 5=pos, 8=pos, 9=neg) ; E. avium if (3=neg, 6=pos, 9=neg, 10=pos, 11=neg). The ILLM algorythm could be a simple way to determine a minimal set of tests for identification of enterococcal strains.

enterococcus; algorithms; bacterial typing

Advances in Experimental Medicine and Biology ; Vol. 418