A review of numerical methods in bacterial identification

A method to implement continuous characters in digital identification keys that estimates the probability of an annotation

PREMISE

Species identification is vital to many disciplines. Digital technology has improved identification tools, but the direct use of characters with continuous states has yet to be fully realized. To achieve full use of continuous characters for identification, I propose a classifier that calculates a posterior probability (degree of belief) in possible name assignments and an estimate of the relative evidence for the candidate annotations.

METHODS

A model for a species is defined using continuous morphological characters, and an algorithm for identification with a naive Bayesian classifier, using the model, is presented. A method of estimating the strength of evidence for candidate species is also described.

RESULTS

The proposed method is applied in two example identifications: native vs. invasive Myriophyllum in North America and vegetative Rhipidocladum bamboos in Mexico. In each instance, the new method provides a probability and estimate of the strength of the probability to enhance the name assignment in situations where taxa are difficult to differentiate using discrete character states.

DISCUSSION

Naive Bayesian classifiers take advantage of the predictive information inherent in continuous morphological characters. Application of this methodology to plant taxonomy advances our ability to leverage digital technology for improved interactive taxonomic identifications.

Non-uniqueness in probabilistic numerical identification of bacteria

In this note we point out an inherent difficulty in numerical identification of bacteria. The problem is that of uniqueness of the taxonomic structure or, in mathematical terms, the lack of statistical identifiability of finite mixtures of multivariate Bernoulli probability distributions shown here.

A Novel 7-Single Nucleotide Polymorphism-Based Clonotyping Test Allows Rapid Prediction of Antimicrobial Susceptibility of Extraintestinal Escherichia coli Directly From Urine Specimens

This prospective study of cellulitis identified β-hemolytic streptococci as the dominating cause in all investigated subgroups. Group C/G streptococci were more frequently detected than group A streptococci. No single clinical feature substantially increased the probability of confirmed streptococcal etiology.

Background. Escherichia coli is a highly clonal pathogen. Extraintestinal isolates belong to a limited number of genetically related groups, which often exhibit characteristic antimicrobial resistance profiles.

Methods. We developed a rapid clonotyping method for extraintestinal E coli based on detection of the presence or absence of 7 single nucleotide polymorphisms (SNPs) within 2 genes (fumC and fimH). A reference set of 2559 E coli isolates, primarily of urinary origin, was used to predict the resolving power of the 7-SNP-based typing method, and 582 representative strains from this set were used to evaluate test robustness.

Results. Fifty-four unique SNP combinations (“septatypes”) were identified in the reference strains. These septatypes yielded a clonal group resolution power on par with that of traditional multilocus sequence typing. In 72% of isolates, septatype identity predicted sequence type identity with at least 90% (mean, 97%) accuracy. Most septatypes exhibited highly distinctive antimicrobial susceptibility profiles. The 7-SNP-based test could be performed with high specificity and sensitivity using single or multiplex conventional polymerase chain reaction (PCR) and quantitative PCR. In the latter format, E coli presence and septatype identity were determined directly in urine specimens within 45 minutes with bacterial loads as low as 10² colony-forming units/mL and, at clinically significant bacterial loads, with 100% sensitivity and specificity.

Conclusions. 7-SNP-based typing of E coli can be used for both epidemiological studies and clinical diagnostics, which could greatly improve the empirical selection of antimicrobial therapy.

Influence of physicochemical and nutritional factors on bacterial diversity in mangrove sediments along the southwest coast of Tamilnadu, India

The present study was undertaken to investigate the diversity of mangrove-associated bacterial genera at Manakudy estuary, Southwest coast of India. The root and rhizosphere sediments of both mangrove and their associated plants were collected from chosen area. Results inferred that the maximum nutrients, THB density, and diversity indices were recorded in rhizosphere and root sample of Avicennia officinalis. Altogether, 258 bacterial strains were isolated from the chosen mangrove samples and screened for nitrogen fixing and phosphate solubilizing ability. Screened result inferred that, 36.43 % isolates had nitrogen fixing and 29.45 % isolates had phosphate solubilizing ability. Here, the genus Bacillus spp. (21.71 %) was the most dominant genera. The bacterial diversity indices, i.e., univariate analysis showed remarkable variation between the chosen samples; however, maximum diversity indices was registered by rhizosphere and root sample of A. officinalis. The 95 % confidence interval and ellipse showed that samples were well mendacious within AvTD and VarTD. Likewise, the multivariate analysis like similarity percentage was good discriminator from 16.64 to 100 % by Bray-Curtis dissimilarity. The prinicipal component analysis (PCA) showed marked variation between the tested bacterial communities. Cluster analysis and non-metric multi-dimensional scaling (MDS) were grouped by Bray-Curtis similarity index which strongly evidence that the rhizosphere and root samples of A. officinalis were highly diversified in the study area.

Influence of probiotic bacterium Bacillus cereus isolated from the gut of wild shrimp Penaeus monodon in turn as a potent growth promoter and immune enhancer in P. monodon

A probiotic bacterium isolated from the gut of wild shrimp Penaeus monodon rendered maximum antagonistic activity against shrimp pathogens and was capable of producing extracellular enzymes. The probiotic bacterium was identified as Bacillus cereus through 16S rRNA sequencing. The lyophilized B. cereus was supplemented with shrimp basal diet at four different concentrations (0.1–0.4%/100 g feed) in D1–D4 diets. The viability of probiotic bacterium in the test diets was evaluated during the study period at various time intervals. The viability ranged from 50.24 ± 1.42 to 180.34 ± 1.30 CFU/g in D1 to D3 diets on the 30th day, whereas it was slightly declined from 45.23 ± 1.30 to 169.13 ± 1.18 CFU/g during the 90th day of storage. A control diet (C), devoid of probiotic supplementation was also simultaneously prepared. During experimentation, P. monodon postlarvae (PL-15) were cultured in individual one tonne capacity FRP tanks in triplicates provided with equal amount of substratum (clay soil) and fed with these respective diets at ad libitum for 90 days. Survival was high (82.0 ± 1.60%) in D4 diet fed shrimp as against a low survival of 65.0 ± 1.33% displayed by control diet fed shrimp. Overall growth responses inferred that a maximum production of 10.45 ± 0.275 g, SGR of 4.40 ± 0.179% and a better FCR of 1.27 ± 0.081 were obtained in D4 diet fed shrimp. However, the water quality parameters showed nonsignificant (P > 0.05) variations among the control and the probiotic treated groups. The tested immunological parameters such as Total haemocyte count, phenoloxidase activity, respiratory burst activity, lysozyme activity, plasma protein concentration and bactericidal activity were higher in D4 diet fed P. monodon, when compared to that of other diets fed shrimp. It is therefore suggested that lyophilized probiotic B. cereus at a concentration of 0.4%/100 g feed was efficient in stimulating the growth and immunity in shrimp.

Identification and typing of Salmonella enterica serotypes isolated from guinea fowl (Numida meleagris) farms in Benin during four laying seasons (2007 to 2010)

The main problem for the local guinea fowl (Numida meleagris) traditional farming and raising system in north-east Benin is the high mortality rate of the keets (up to 70%) due to a combination of climatic, nutritional, hygienic and infectious causes. The present study was carried out to identify and compare the isolates of Salmonella enterica from necropsied keets, laying guinea fowl, surrogate hen mothers, other contact animal species and farmers during four laying seasons (2007 to 2010). S. enterica belonging to eight different serotypes (Adelaide, Farakan, Kingston, Legon, Luke, Oakland, Sangalkam and Teshie) and one untypable isolate were isolated from 13 to 19% of the necropsied keets. The serotypes Adelaide, Farakan, Luke, Sangalkam and Teshie and the untypable isolate were isolated in only one township during 1 year of sampling, while serotypes Oakland, Legon and Kingston were present in two to three townships for 2 to 3 years of sampling. Serotypes Farakan, Kingston, Legon, Oakland and Sangalkam were also isolated from faecal samples of laying guinea fowl and/or surrogate domestic fowl hen mothers. Further comparison by pulsed-field gel electrophoresis and virulotyping provided evidence for their clonality within each of those five serotypes and therefore for the adult guinea fowl and/or hens as the most probable origin of contamination of the keets. The antibiotic resistance profiles, with all isolates resistant to oxacillin, sulfamethoxazol and colistin, emphasize the rise of antibiotic resistance in salmonellas from guinea fowl in this area and the need for alternative therapy policies for these birds.

Typing of O26 enterohaemorrhagic and enteropathogenic Escherichia coli isolated from humans and cattle with IS621 multiplex PCR-based fingerprinting

AIMS

This study evaluated a typing method of O26:H11 enterohaemorrhagic and enteropathogenic Escherichia coli (EHEC and EPEC) based on the variation in genomic location and copy numbers of IS621.

METHODS AND RESULTS

Two multiplex PCRs, targeting either the left (5') or right (3') IS/chromosome junction of 12 IS621 insertion sites and one PCR specific of another truncated copy, were developed. Thirty-eight amplification profiles were observed amongst a collection of 69 human and bovine O26:H11 EHEC and EPEC. Seventy-one per cent of the 45 EHEC and EPEC with identical IS621 fingerprints within groups of two, three or four isolates had >85% pulsed field gel electrophoresis (PFGE) profile similarity, including four groups of epidemiologically related EHEC or EPEC, while most of the groups had <85% similarity between each others. Epidemiologically related EHEC from each of three independent outbreaks in Japan and Belgium also exhibited identical IS621 fingerprints and PFGE profiles.

CONCLUSIONS

The IS621 fingerprinting and the PFGE are complementary typing assays of EHEC and EPEC; though, the former is less discriminatory.

SIGNIFICANCE AND IMPACT OF THE STUDY

The IS621 printing method represents a rapid (24 h) first-line surveillance and typing assay, to compare and trace back O26:H11 EHEC and EPEC during surveys in farms, multiple human cases and outbreaks.

Fluorescent amplified fragment length polymorphism probabilistic database for identification of bacterial isolates from urinary tract infections

The ability of the fluorescent amplified fragment length polymorphism (FAFLP) technique to identify bacterial isolates from urinary tract infections (UTIs) was investigated. FAFLP was carried out using the single primer combination MseI plus CT and EcoRI plus 0, and information-rich FAFLP profiles were generated from all 69 UTI isolates studied, which comprised both gram-negative and gram-positive bacteria encompassing eight genera. The genetic relatedness of these 69 bacteria was determined by cluster analysis, and this revealed eight main groups corresponding to the eight bacterial genera. Finer discrimination on the same dendrogram showed species and subspecies differentiations, thus demonstrating the potential of FAFLP for describing a wide diversity range within microbial populations. The interpretation of FAFLP profiles is often complicated because it relies upon the investigator interpreting dendrograms; this process may be subjective if the tree is complicated, particularly if it includes polytomies (unresolved nodes). Therefore, we have developed a method based on Bayes' theorem for the identification of bacteria against an FAFLP probabilistic identification matrix. Thus, FAFLP is suitable for the objective identification of causal agents of UTI, and the procedure offers great potential in the clinical laboratory.

Numerical evaluation of minimal biochemical test combinations for the identification of Enterobacteriaceae species

Species identification of 572 clinical isolates of Enterobacteriaceae was determined using an in-house biochemical test panel and API 20E. The overall agreement was 82.7 per cent. When a numerical identification programme for Enterobacteriaceae was used to analyze the results of 440 E.coli isolates, the agreement with API 20E increased from 81.4 to 91.1 per cent. Numerical identification was then employed to evaluate 14 routine biochemical test panels used in 12 Swedish laboratories, and an additional 18 potential new test panel variants. The theoretical identification power of the panels was measured by a scoring system, based on simulated test results for the 14 most common species of Enterobacteriaceae. The results showed wide variations in the performance of different panels, with a number of panels being of insufficient quality. Among the 32 test panels, 6 panels indicated excellent performance. These should be further evaluated for routine laboratory performance. Numerical identification methods constitute powerful tools in the hands of the clinical microbiologist to aid in species identification as well as to analyze the quality of identification charts in use.

Bayesian predictiveness, exchangeability and sufficientness in bacterial taxonomy

We present a theory of classification and predictive identification of bacteria. Bacterial strains are characterized by a binary vector and the taxonomy is specified by attaching a label to each vector. The theory is developed from only two basic assumptions, viz. that the sequence of pairs of feature vectors and the attached labels is judged (infinitely) exchangeable and predictively sufficient. We derive expressions for the training error and the probability of identification error and show that latter is an affine function of the former. We prove the law of large numbers for identification matrices, which contain the fundamental information of bacterial data. We prove the Bayesian risk consistency of the predictive identification rule given by the theory and show that the training error is a consistent estimate of the generalization error.

Advances in the bacteriology of the coliform group: their suitability as markers of microbial water safety

Advances in the elaboration of novel genomic types of beta-galactosidase-positive Enterobacteriaceae and comprehensive studies of their habitats have resulted in an innovative approach to the assessment of the merits and shortcomings of the thermotrophic and fecal species Escherichia coli and all other coliforms as markers of the microbiological safety of water. As one of the consequences, it is recommended to abolish the "technical" designation fecal coliforms because their current method of detection will result in the isolation of thermotrophic organisms that have been demonstrated, beyond a doubt, to be of environmental, rather than uniquely enteric origin. Additional population studies have demonstrated that none of the coliforms can function as reliable markers for all enteric pathogens (index organisms sensu Ingram), nor be of use in validating adequate processing for safety of raw water, which represents the indicator function of markers, as defined by Ingram. Future studies along these lines will have to provide the data required to assess the suitability of additional markers for the reliable monitoring of drinking water for microbiological safety.

Miniaturized tests for computer-assisted identification of motile Aeromonas species with an improved probability matrix

AIMS

To develop miniaturized tests for the phenotypic identification of motile Aeromonas species using an improved probability matrix.

METHODS AND RESULTS

Conventional tests were miniaturized for use in 96-well plates, and their performance assessed using 60 aeromonads comprising type and reference strains as well as clinical, fish and water isolates. A revised probability matrix for Aeromonas hybridization groups 1-14, including A. allosaccharophila, A. bestiarum, A. encheleia and A. popoffii, was developed. Using 26 tests, all the reference strains were correctly identified with the revised probability matrix, and 80% of the isolates were correctly identified at a Willcox probability level of 95%.

CONCLUSION

The compact test format, coupled with a robust identification matrix, provides a convenient basis for identifying motile aeromonads.

SIGNIFICANCE AND IMPACT OF THE STUDY

The identification system for identifying aeromonads will be of use to medical and veterinary laboratories undertaking disease diagnosis.

Minimizing stochastic complexity using local search and GLA with applications to classification of bacteria

In this paper, we compare the performance of two iterative clustering methods when applied to an extensive data set describing strains of the bacterial family Enterobacteriaceae. In both methods, the classification (i.e. the number of classes and the partitioning) is determined by minimizing stochastic complexity. The first method performs the minimization by repeated application of the generalized Lloyd algorithm (GLA). The second method uses an optimization technique known as local search (LS). The method modifies the current solution by making global changes to the class structure and it, then, performs local fine-tuning to find a local optimum. It is observed that if we fix the number of classes, the LS finds a classification with a lower stochastic complexity value than GLA. In addition, the variance of the solutions is much smaller for the LS due to its more systematic method of searching. Overall, the two algorithms produce similar classifications but they merge certain natural classes with microbiological relevance in different ways.

Classification of Enterobacteriaceae by minimization of stochastic complexity

A new method for classifying bacteria is presented and applied to a large set of biochemical data for the Enterobacteriaceae. The method minimizes the bits needed to encode the classes and the items or, equivalently, maximizes the information content of the classification. The resulting taxonomy of Enterobacteriaceae corresponds well to the general structure of earlier classifications. Minimization of stochastic complexity can be considered as a useful tool to create bacterial classifications that are optimal from the point of view of information theory.

Pyrolysis mass spectrometry: data processing in classification studies

A scheme for numerical processing of pyrolysis mass spectrometry (Py-MS) data is detailed, along with methods for combining the results of conventional phenotypic and Py-MS taxonomic surveys. The importance of this combined approach in polyphasic taxonomy is emphasised; it yields data on cell composition and the nutritional and physiological interactions of strains with their environment. Large collections of strains can be surveyed rapidly and economically, yielding presumptive classifications, which may then be confirmed with a few representative strains in more demanding, difficult and expensive approaches. An objective, non-arbitrary method of establishing suitable cut-off points to delineate group structures in dendrograms is also described.

Re-speciation of the original reference strains of serovars in the Citrobacter freundii (Bethesda-Ballerup group) antigenic scheme of West and edwards

The antigenic scheme for the Bethesda-Ballerup group of bacteria established by West and Edwards in 1954 has continued to be applied as a serotyping scheme for Citrobacter freundii. In 1993, however, the classification of the Citrobacter was drastically revised and the species C. freundii redefined by Brenner et al. Accordingly, to judge the propriety to continuously use a single antigenic scheme for the C. freundii complex, the 90 reference strains listed in the antigenic scheme for C. freundii by West and Edwards were characterized phenotypically and specified based on the revised classification. Of these 90 strains, two strains of Hafnia alvei and one of Escherichia coli were found. Among the remaining 87 reference strains, Citrobacter youngae was the predominant species (40 strains), followed by Citrobacter braakii (25 strains), Citrobacter werkmanii (13 strains), and the unnamed Citrobacter genospecies 10 of Brenner et al (six strains). Citrobacter freundii, as redefined, accounted for only three strains and ranked behind the other four species. No overlapping with most of the 42 O-groups and 82 H-antigens was recognized between species with few exceptions. O-groups 1-9 inclusive, which were estimated to represent more than 90% of the former C.freundii strains, occurred in strains of C. youngae and C. braakii; and all nine strains of O-group 29, formerly known as the Ballerup group, were identified as C. braakii. These findings suggest that further study of the serotyping system is needed for all H2S-producing Citrobacter species.

Non-uniqueness in probabilistic numerical identification of bacteria

In this note we point out an inherent difficulty in numerical identification of bacteria. The problem is that of uniqueness of the taxonomic structure or, in mathematical terms, the lack of statistical identifiability of finite mixtures of multivariate Bernoulli probability distributions shown here.

Software tools and databases for bacterial systematics and their dissemination via global networks

The dynamic expansion of the taxonomic knowledge base is fundamental to further developments in biotechnology and sustainable conservation strategies. The vast array of software tools for numerical taxonomy and probabilistic identification, in conjunction with automated systems for data generation are allowing the construction of large computerised strain databases. New techniques available for the generation of chemical and molecular data, associated with new software tools for data analysis, are leading to a quantum leap in bacterial systematics. The easy exchange of data through an interactive and highly distributed global computer network, such as the Internet, is facilitating the dissemination of taxonomic data. Relevant information for comparative sequence analysis, ribotyping, protein and DNA electrophoretic pattern analysis is available on-line through computerised networks. Several software packages are available for the analysis of molecular data. Nomenclatural and taxonomic 'Authority Files' are available from different sources together with strain specific information. The increasing availability of public domain software, is leading to the establishment and integration of public domain databases all over the world, and promoting co-operative research projects on a scale never seen before.

Identification and typing of Proteus penneri and Proteus vulgaris biogroups 2 and 3, from clinical sources, by computerized analysis of electrophoretic protein patterns

Seventy-six strains of the Proteus vulgaris complex (Pr. penneri and Pr. vulgaris biogroups 2 and 3) were characterized by one-dimensional SDS-PAGE of cellular proteins. The protein patterns were highly reproducible. The strains came from various countries and were mainly of human origin: urine (28), respiratory tract (13), wounds (8), faeces (7), blood (3), miscellaneous sources (6) and unknown sources (11). The patterns of these strains, together with those of the type strains of seven Morganella, Proteus and Providencia species were subjected to two numerical analyses. In the first, in which the principal protein bands (in the 35.0-42.0 kDa range) were excluded, the strains of the Pr. vulgaris complex formed four clusters at the 83% similarity level. These corresponded to Pr. penneri, Pr. vulgaris biogroup 2, and two clusters (3a and 3b) represented biogroup 3. Each of these clusters was distinct from the Morganella, Proteus and Providencia reference strains. In the second analysis, which included all the protein bands, the 41 Pr. penneri strains showed little heterogeneity but 17 subphenons could be recognized among the 35 strains of Pr. vulgaris biogroups 2 and 3. These results support the division of biogroup 3 strains into at least two separate taxa. Other results indicate that biogroup 3 is heterogeneous and may contain further genomic groups. The method also provides a basis for typing clinical strains of Pr. vulgaris biogroups 2 and 3.

Simplified techniques for identifying foodborne yeasts

Four problematic areas associated with the identification of foodborne yeasts are discussed. These consist of (1) the inability of conventional identification tests to recognize some common and important foodborne yeasts characterized by genomic differences (e.g., Saccharomyces cerevisiae, S. bayanus and S. pastorianus); (2) the delay in application of non-traditional identification methods such as DNA fingerprinting, chromosome karyotyping, protein electrophoretic patterns and fatty acid profiles for routine identification purposes; (3) the lack of commercially available manual or automated identification systems dedicated to the diagnosis of foodborne yeasts; and (4) the disregard for considering ecological frequency of yeasts in computerized probabilistic identification systems.

Microbiological characterization of a fuel-oil contaminated site including numerical identification of heterotrophic water and soil bacteria

Seven soil samples and seven groundwater samples from a site contaminated with fuel-oil were investigated using several chemical and microbiological techniques. In soil samples, 500 to 7,500 mg/kg of total hydrocarbons were found. These samples contained no n-alkanes but iso- and branched chain alkanes. No polychlorinated biphenyls could be detected. Microbiological investigations included estimations of total cell counts, viable cell counts on different media, and numbers of methylotrophic, denitrifying, sulphate reducing, anaerobic (with the exception of methanogenic organisms), and hydrocarbon degrading bacteria. Viable and hydrocarbon degrading bacteria were found in all samples. A total of 1,366 pure cultures was characterized morphologically and physiologically and identified by numerical identification using a data base of more than 4,000 reference strains. Groundwater samples were dominated by gram-negative bacteria of the generaPseudomonas, Comamonas, Alcaligenes, andAcinetobacter, which were also found in soil samples. In addition, more grampositive bacteria belonging to the generaArthrobacter, Nocardia, andBacillus could be isolated from soil samples.

Typing of Proteus mirabilis from clinical sources by computerized analysis of electrophoretic protein patterns

Seventy-five strains of Proteus mirabilis (mainly of human origin) were characterized by one-dimensional SDS-PAGE of cellular proteins. The strains came from various countries; 24 were from urine, 17 from faeces (including one from a dog and one from a monkey), 12 from the respiratory tract (including nine from sputum), four from a cerebral abscess, 12 from miscellaneous sources and six from unknown sources. The protein patterns, which contained 45 to 50 discrete bands, were highly reproducible. The patterns of the 75 P. mirabilis cultures plus those of the type strains of six Proteus and Providencia species were used as the basis for two numerical analyses. In the first, which included all the protein bands, the P. mirabilis cultures formed nine clusters at the 85% S level. In the second analysis, in which the principal protein bands (in the 34.0-44.6 kDa range) were excluded, 74 of the 75 cultures of P. mirabilis formed a single cluster at the 77% S level distinct from the six Proteus and Providencia reference strains. The P. mirabilis strain which failed to cluster with the others had a background band pattern typical of the species in the lower molecular weight region but appeared to be less typical in the heavier bands. It is concluded that high resolution PAGE combined with computerized analysis of protein patterns provides the basis for typing clinical strains of P. mirabilis. Reference strains of each of the nine PAGE types identified are available from NCTC for inclusion in future studies.

Numerical analysis of electrophoretic protein patterns of Morganella morganii strains from faeces, wound, urine and other clinical sources

Sixty-five strains of Morganella morganii (mainly of human origin) were characterized by one-dimensional SDS-PAGE of cellular proteins. The strains came from various countries; 13 were from stools (including one from a toucan), 13 from wounds, 11 from urine, five from blood (including one from a snake), five from the respiratory tract (four sputum, one lung), 12 from miscellaneous sources and six from unknown sources. The protein patterns, which contained 45 to 50 discrete bands, were highly reproducible. The patterns of 67 M. morganii cultures plus those of the type strains of seven Proteus and Providencia species were used as the basis for two numerical analyses. In the first, which included all the protein bands, the M. morganii strains formed 21 clusters at the 91% S level. In the second analysis, in which the principal protein bands (in the 31.6-43.2 kDa range) were excluded, the 67 M. morganii cultures formed a single cluster at the 80% S level distinct from the seven Proteus and Providencia reference strains. We conclude that high resolution PAGE combined with computerized analysis of protein patterns provides the basis for typing clinical strains of M. morganii. Reference strains of each of the 21 PAGE types identified are available from NCTC for inclusion in future studies.

Numerical analysis of electrophoretic protein patterns of Providencia stuartii strains from urine, wound and other clinical sources

Eighty-six strains of Providencia stuartii (mainly of human origin) were characterized by one-dimensional SDS-PAGE of cellular proteins. The strains came from various countries; 52 were from urine, 11 from wounds, five from blood (one of these also from urine), four from ear infections, two each from faeces and sputum, one from 'alimentation' and nine from unknown sources. The protein patterns, which contained 45 to 50 discrete bands, were highly reproducible. The patterns of 46 Prov. stuartii strains (selected to represent the full range of protein pattern diversity) plus those of the type strains of the four other Providencia species were used as the basis for two numerical analyses. In the first, which included all the protein bands, the Prov. stuartii strains formed 13 clusters at the 88% S level. In the second analysis, in which the principal protein bands (in the 33.8-40.7 kDa range) were excluded, 45 of the 46 Prov. stuartii strains formed a single cluster at the 82% S level, whilst the four Providencia reference strains remained unclustered. The 40 strains of Prov. stuartii not included in the cluster analysis were assigned to a protein type by calculating their similarity with the strains in the database used for the cluster analysis. We conclude that high resolution PAGE combined with computerized analysis of protein patterns provides the basis for typing clinical strains of Prov. stuartii. Reference strains of each of the 13 PAGE types identified are available from NCTC for inclusion in future studies.

Development of an expert system for bacterial identification: study of a prototype for identifying beta-galactosidase positive enterobacteria

A prototype of an expert system for the identification of beta-galactosidase positive Enterobacteriaceae has been developed, for use with the API 20 EC kit. The system is implemented in Prolog on an IBM PC AT with 640 K of central memory and 20 megabytes of secondary memory. Its objectives are to highlight errors that can occur when the kit is in use. It can indicate the presence of new groups or species and give advice or suggest additional tests for the differentiation of the new species from those included in the kit.

Numerical analysis of electrophoretic protein patterns of Providencia rettgeri strains from human faeces, urine and other specimens

Thirty-one strains of Providencia rettgeri (mainly from humans) were characterized by one-dimensional SDS-PAGE of cellular proteins. The strains came from various countries and comprised 14 from urine, eight from faeces, two from bile (plus one from the liver of a sheep), two from sputum, one from an insect pupa and three the sources of which were unknown. Also included, for reference purposes, were the type strains of the four other Providencia species. The protein patterns, which contained 45-50 discrete bands, were highly reproducible and were used as the basis for two numerical analyses. In the first, which included all the protein bands, the 31 Prov. rettgeri strains formed 13 clusters at the 84% S level. In the second analysis, in which the principal protein bands (in the 33.3-41.3 kD range) were excluded, 29 of the 31 Prov. rettgeri strains formed a single cluster at the 81% S level, whilst the four Providencia reference strains remained unclustered. We conclude that high resolution PAGE combined with computerized analysis of protein patterns provides the basis for typing clinical strains of Prov. rettgeri. Reference strains of each of the 13 PAGE types identified are available from NCTC for inclusion in future studies.

Evaluation of the Titertek-NF system for identification of gram-negative nonfermentative and oxidase-positive fermentative bacteria

The Titertek-NF (TT-NF) system (Flow Laboratories GmbH, Meckenheim, Federal Republic of Germany) was evaluated for the identification of 1,289 strains of gram-negative, nonfermentative bacteria and some gram-negative, oxidase-positive bacteria. The oxidase test was also performed. Identifications were classified as correct, not identified (two or more taxa possible and identification score of less than 80%; supplementary tests for furthering the identification were not performed), and incorrect. Correct identification results were further subdivided by the correct level of species or biotype identification as greater than or equal to 98% (category a), 90 to 97% (category b), and 80 to 89% (category c). When compared with conventional identification results, the TT-NF system correctly identified 90.3% of strains (1,164 of 1,289 strains), with 72.5% (935 strains) belonging to category a, 14.7% of strains (189 strains) belonging to category b, and 3.1% of strains (40 strains) belonging to category c. Among the remaining strains, 104 (8.1%) were not identified and 14 (1.1%) were misidentified, and the system failed to generate identification results for 7 strains (0.5%). Reactions within the TT-NF system were reproducible, with an estimated probability of erroneous test results of 0.2%.

The effect of cooling rate, freeze-drying suspending fluid and culture age on the preservation of Campylobacter pylori

The effects of freezing rate, suspending fluid and age of culture on the ability of four strains of Campylobacter pylori to survive and recover from freeze-drying were examined. Freeze-drying by standard procedures generally resulted in an overall loss in viability of between 3 and 7 log units. The exact cause of poor recovery by C. pylori was not established but strain differences were detected, with NCTC 11637 (type strain) surviving better than NCTC 11638 and NCTC 11639. Recovery of the poorest growing strain (NE 26695) was notably more erratic. The largest loss in viability occurred at the primary drying stage. Losses resulting from freezing and secondary drying were less marked and the rate of freezing had only a marginal effect on recovery. Nineteen different freeze-drying suspending fluids were investigated. Overall the best recovery results were obtained with 5% inositol-broth (or horse serum) plus 25% glucose, at pH 7.0, in which loss of viability was typically about 4 log units. Other factors, such as age of culture and number of viable bacteria in the before-dry suspension, did not have a significant effect on survival. We conclude from these results that C. pylori can survive freeze-drying, albeit in small numbers, but the degree of recovery is apparently largely strain dependent.

Numerical analysis of electrophoretic protein patterns of Providencia rustigianii strains from human diarrhoea and other sources

Twenty strains of Providencia rustigianii (including the type strain of Prov. friedericiana) have been characterized by one-dimensional SDS-PAGE of cellular proteins. They comprised 12 strains (almost exclusively associated with the intestinal tract) from humans, plus eight largely from the intestinal tract of pig, penguin and environmental sources. The protein patterns, which contained 45-50 discrete bands, were highly reproducible and were used as the basis for two numerical analyses. In the first, which included all the protein bands, the 20 Prov. rustigianii strains formed six clusters at the 88% S level. One of these clusters included the type strains of both Prov. friedericiana and Prov. rustigianii, thereby confirming the synonymy of these two species. In the second analysis, the principal protein bands were excluded. At the 86% S level the 20 Prov. rustigianii strains formed a single cluster, whilst a field strain of Morganella morganii and the respective type strains of three other Providencia species remained unclustered. The total protein pattern of the type strain of Prov. alcalifaciens was very similar to that of Prov. rustigianii phenon 3 and the M. morganii field strain, which indicates that careful biochemical characterization may be necessary to ascribe strains to a species before typing by the PAGE technique. Alternatively, a selective analysis of the protein bands may be used to confirm the identity of the strains, as shown in this study.

Evaluation of Mast-ID 15 system for identifying Enterobacteriaceae, some Vibrionaceae, and Acinetobacter

Six hundred and twenty one strains (555 Enterobacteriaceae, 46 Vibrionaceae, and 20 Acinetobacter) were examined in the Mast system. The results were consulted in the code book supplied by the manufacturer and those not listed were processed through the manufacturer's full database held on an Apricot microcomputer in our laboratory. The proportion of strains correctly identified was 88%, with 9% not identified, and 3% incorrectly identified.

Identification and distribution of Flavobacterium meningosepticum in clinical material

During the 19 year period ending December 1984, 82 (1.7%) of 4840 strains of Gram-negative non-fermentative bacteria submitted to the National Collection of Type Cultures (NCTC) for computer-assisted identification were Flavobacterium meningosepticum. These figures suggest either that F. meningosepticum occurs only rarely in clinical material in the UK, or that when it does occur it is not always recognized. The sources from which the NCTC strains were isolated are reported and also the characteristics by which the species may be identified. The clinical significance of F. meningosepticum is reviewed, particularly its role in neonatal meningitis, of which there have been but two reported cases in the UK, both imported. The susceptibility of the species to antimicrobial agents is discussed and also antimicrobial therapy of neonatal meningitis due to this species.

A probability matrix for the identification of species of Vibrio and related genera

A matrix for the probabilistic identification of species of Vibrio and related genera has been constructed using the data from 1091 strains collected throughout the world and classified. Thirty-eight phenons are included in the matrix, 31 of these represent previously identified species or biovars and seven represent phenons which could not be identified and may represent new species. The identification matrix incorporates 81 characters although a subset of 30 tests can be used to distinguish the 38 phenons from each other. The additional 51 tests were included to assist the identification of some strains for which the initial 30 tests were inadequate. No significant cluster overlap was found at the 5% level and the identification score for the Hypothetical Median Organism of each cluster exceeded 0.9999 in all cases.

The identification of Pseudomonas cepacia and its occurrence in clinical material

During the 19 year period ending December 1984, 4840 strains of Gram-negative non-fermentative bacteria were submitted to the National Collection of Type Cultures for identification. Of these, 195 strains (4.0% of the total) were identified as Pseudomonas cepacia which demonstrates both that the species is regularly encountered in clinical material in the UK and that several laboratories have experienced difficulty in identifying the organism. The sources from which the 195 strains were isolated are reported and also the characteristics by which the species may be recognized. The clinical significance of Ps. cepacia is reviewed, and the resistance of this species to disinfectants and antimicrobial agents commonly used to treat pseudomonas infections is discussed to underline the necessity for the precise identification of Ps. cepacia.

Identification and distribution of Pseudomonas stutzeri in clinical material

During the 19 year period ending December 1984, 114 (2.4%) of 4840 strains of Gram-negative non-fermentative bacteria submitted, mainly by laboratories in the UK, to the National Collection of Type Cultures for computer-assisted identification were strains of Pseudomonas stutzeri. These figures suggest that Ps. stutzeri is a relatively uncommon species in clinical material in the UK but that when it does occur laboratories have difficulty in identifying it. The sources from which the strains were isolated and also characteristics of the species by which it may be recognized are reported. The clinical significance of Ps. stutzeri is discussed and also the susceptibility of this species to antimicrobial agents.

Misuse and interlaboratory test reproducibility of API 20E system

One hundred strains were referred to us for identification because they apparently could not be identified satisfactorily with the API 20E system (appareils et procédés d'identification). The inability to identify 31 strains was due primarily to failure to follow the manufacturer's instructions. Twenty six further strains were found to have been correctly identified by the sender's own API 20E results, so that only the remaining 43 strains definitely fell into the category for which our identification service was intended. Eighteen of the 43 strains not identified by the sender were identified by us using the API 20E system, and several possible reasons are given to explain the differences in these results. The remaining 25 strains either could not be identified by us on the API 20E system or, in the case of 13, they could not be identified by our conventional system and therefore no comparison could be made. The average interlaboratory probability of errors for the API 20E tests was 6.1%.

A probability matrix for the identification of vibrios

A probability matrix for computer-assisted identification of vibrios has been constructed, based on the API 20E system. Data were gathered from 173 strains representing 31 taxa of vibrios and related organisms, from a variety of sources. The matrix was tested internally by four statistical programs. Program OVERMAT tested the separation and program MOSTTYP the discretion and homogeneity of the taxa. Most of the taxa were satisfactory but a few were less so; reasons for this are discussed. Program CHARSEP and program DIACHAR tested the separation and diagnostic values, respectively, of the characters used. The overall test error was 4.5%. The matrix was assessed externally by its performance in the identification of vibrio-like strains isolated from freshwater. Of 243 wild strains, 79.4% were identified with ten taxa, with a Willcox score of greater than or equal to 0.99.

A general-purpose system for characterizing medically important bacteria to genus level

A computer program and accompanying data matrix have been prepared for bacteria of medical interest, to assist the assignment of an unidentified bacterium to the most likely genus. The results on a set of relatively simple tests are entered. The program prints the more likely genera, followed by a list of diagnostic tables in Cowan & Steel (1974) and Buchanan & Gibbons (1974). Where available, identification matrices for further computer-assisted study, are presented. This program may be of particular help in laboratories where a wide range of bacteria have to be identified.