Bacterial Identification, first described by Bryant (1986), is a program that provides a novel way of enhancing the teaching of systematic bacteriology and numerical identification procedures. A student is assigned an unknown isolate from a list of bacteria. The student's objective is to identify the unknown isolate using the least number of tests from the set of tests available. Simultaneously the computer tries to identify the unknown using optimised and random selection of tests. A student can compare their progress against that of the computer. The program has been improved and the data matrix on which it was based has been revised by the addition of 14 more species, an additional test and by updating the original probability matrix. The program, Bacterial Identification, is available as 'freeware' and has been placed with organisations that distribute such software.

A program has been written to help identify the correct RKC codes (Rogosa, Krichevsky and Colwell, 1986) for describing microbiological attributes. Terms, which may be complete words, parts of words, or numbers, can be typed into the computer. A search is made of the set of about 13,800 RKC codes and statements that describe them. Statements that contain the search terms are displayed in a scrolling window so that the required statement(s) can be identified. A list of selected RKC statements can be built up in a second window for subsequent use by other software packages. The program called RKCLIST runs under the MS-DOS operating system. The software routines used to create RKCLIST have been written so that they can be incorporated into other software packages that make use of the RKC coding scheme.

An investigation was carried out to assess the feasibility of using an expert systems approach to assist in the identification of unknown isolates of bacteria. A system was developed using Lisp which utilized the knowledge stored in standard bacteriological texts. A comparison of the expert systems approach and the probabilistic approach based on Bayes Theorem was made together with the advantages and disadvantages of each approach.

A number of programs are described for the development and evaluation of probabilistic identification matrices for use with computer-assisted identification. The program BEST reads an initial matrix of per cent probabilities for all the binary characters examined during a cluster analysis and determines the most useful set of tests for distinguishing taxa in the matrix. Program RESORT creates from the initial matrix an identification matrix in which the order of the tests may be different or the number of tests reduced. A printed version of a matrix can be produced by MATPRINT, which creates tables giving the per cent probability of a positive test result and the test results presented as '-', 'v' and '+' depending on a user-specified threshold. Program IDSC evaluates an identification matrix by calculating the best identification score for each taxon in the matrix using the expected result for each test. The programs were written in FORTRAN 77 and can be run on any micro-computer using the PC/MS-DOS operating system.

Three programs are described for evaluating and characterising data collected during numerical taxonomic studies of bacteria. The program VARIANCE compares replicate cultures and evaluates the reproducibility of each character. Also it identifies those characters that should be excluded from subsequent taxonomic analysis because of their poor reproducibility. GPROPS summarises the properties of clusters of strains that have been defined from a cluster analysis, it can produce a probabilistic identification matrix and compares each strain within a cluster with the Hypothetical Mean Organism (HMO) of that cluster. OVCLUST is an implementation of the program described by Sneath (1979) which calculates overlap statistics between major clusters. These programs are designed to complement the CLUSTAN package (Wishart, 1982) which is often used for cluster analysis of bacterial taxonomic data. The programs were written in FORTRAN 77 and implemented on an IBM PC using MS-DOS.

A microcomputer based system for the identification of unknown isolates of Bacillus species is described. The identification matrix includes 78 test probabilities for 38 recognised species and other groups in the genus Bacillus and it is based on the work of Logan and Berkeley (1984). Morphological characters together with the results of tests using API 20E and API 50CHB, read after 24 and 48 h incubation, are used to obtain a probabilistic identification of an unknown aerobic endospore forming rod. Any differences between the observed and expected results for any identified organism are listed. Identification can be attempted on the basis of a limited set of test results, although this is rarely if ever done with this largely API based system, and if the unknown cannot be successfully identified then a set of additional tests can be selected which should permit identification. The computer system can store and recall test results entered for any isolate. This feature allows the accumulation of data on isolates which could be used to update the identification matrix in future taxonomic studies.