[Selection of phenotype characteristics in differentiation of various species of enterococci using an inductive mathematical program].

In this paper a novel computer example-based learning system (Inductive learning by logic minimization) was used to determine the sufficient set of biochemical reactions and necessary conditions that have to be fulfilled for the correct identification of enterococci isolated from human specimens. Several combinations for accurate identification of Enterococcus faecalis and Enterococcus feacium from other enterococci were found. The simplest combination set for E. faecalis identification consists of 3 reactions: growth in 0.04% K-tellurite, mannitol acidification, failure to acidify raffinose. The simplest set for E. faecium differentiation from other enterococcal species is performed by the following 3 reactions: failure to acidify D-xylose and Na-pyruvate, and acidification of L-arabinose. Only 4 tests are sufficient (acid from D-xylose, mannitol, L-arabinose and Na-pyruvate) for the correct identification of these two, the most frequently isolated species. If it is necessary to analyze all enterococcal species, then even three combinations of 9 tests for the correct differentiation of any of them, are available.