Machine-Learning Model for Prediction of Cefepime Susceptibility in Escherichia coli from Whole-Genome Sequencing Data.

The declining cost of performing bacterial whole-genome sequencing (WGS) coupled with the availability of large libraries of sequence data for well-characterized isolates have enabled the application of machine-learning (ML) methods to the development of nonlinear sequence-based predictive models. We tested the ML-based model developed by Next Gen Diagnostics for prediction of cefepime phenotypic susceptibility results in Escherichia coli. A cohort of 100 isolates of E. coli recovered from urine (n = 77) and blood (n = 23) cultures were used. The cefepime MIC was determined in triplicate by reference broth microdilution and classified as susceptible (MIC of ≤2 µg/mL) or not susceptible (MIC of ≥4 µg/mL) using the 2022 Clinical and Laboratory Standards Institute breakpoints. Five isolates generated both susceptible and not susceptible MIC results, yielding categorical agreement of 95% for the reference method to itself. Categorical agreement of ML to MIC interpretations was 97%, with 2 very major (false, susceptible) and 1 major (false, not susceptible) errors. One very major error occurred for an isolate with blaCTX-M-27 (MIC mode, ≥32 µg/mL) and one for an isolate with blaTEM-34 for which the MIC cefepime mode was 4 µg/mL. One major error was for an isolate with blaCTX-M-27 but with a MIC mode of 2 µg/mL. These preliminary data demonstrated performance of ML for a clinically important antimicrobial-species pair at a caliber similar to phenotypic methods, encouraging wider development of sequence-based susceptibility prediction and its validation and use in clinical practice.

Performance of the Reveal Rapid Antibiotic Susceptibility Testing System on Gram-Negative Blood Cultures at a Large Urban Hospital.

Timely and effective antibiotic treatment is vital for sepsis, with increasing incidence of antimicrobial-resistant bacteremia driving interest in rapid phenotypic susceptibility testing. To enable the widespread adoption needed to make an impact, antibiotic susceptibility testing (AST) systems need to be accurate, enable rapid intervention, have a broad antimicrobial menu and be easy to use and affordable. We evaluated the Specific Reveal (Specific Diagnostics, San Jose, CA) rapid AST system on positive blood cultures with Gram-negative organisms in a relatively resistant population in a large urban hospital to assess its potential for routine clinical use. One hundred four randomly selected positive blood cultures (Virtuo; bioMérieux) were Gram stained, diluted 1:1,000 in Pluronic water, inoculated into 96-well antibiotic plates, sealed with the Reveal sensor panel, and placed in the Reveal instrument for incubation and reading. The MIC and susceptible/intermediate/resistant category was determined and compared to results from Vitek 2 (bioMérieux) for the 17 antimicrobials available and to Sensititre (Thermo Fisher) for 24 antimicrobials. Performance was also assessed with contrived blood cultures with 33 highly resistant strains. Reveal was in 98.0% essential agreement (EA) and 96.3% categorical agreement (CA) with Sensititre, with just 1.3% very major error (VME) and 97.0%/96.2%/1.3% EA/CA/VME versus Vitek 2. Reveal results for contrived highly resistant strains were equivalent, with EA/CA/VME of 97.7%/95.2%/1.0% with CDC/FDA Antibiotic Resistance Isolate Bank references. Average time to result (TTR) for Reveal was 4.6 h. Sample preparation was relatively low skill and averaged 3 min. We conclude that the Reveal system enables accurate and rapid susceptibility testing of Gram-negative blood cultures.

Evaluation of a fully automated bioinformatics tool to predict antibiotic resistance from MRSA genomes.

OBJECTIVES: The genetic prediction of phenotypic antibiotic resistance based on analysis of WGS data is becoming increasingly feasible, but a major barrier to its introduction into routine use is the lack of fully automated interpretation tools. Here, we report the findings of a large evaluation of the Next Gen Diagnostics (NGD) automated bioinformatics analysis tool to predict the phenotypic resistance of MRSA. METHODS: MRSA-positive patients were identified in a clinical microbiology laboratory in England between January and November 2018. One MRSA isolate per patient together with all blood culture isolates (total n = 778) were sequenced on the Illumina MiniSeq instrument in batches of 21 clinical MRSA isolates and three controls. RESULTS: The NGD system activated post-sequencing and processed the sequences to determine susceptible/resistant predictions for 11 antibiotics, taking around 11 minutes to analyse 24 isolates sequenced on a single sequencing run. NGD results were compared with phenotypic susceptibility testing performed by the clinical laboratory using the disc diffusion method and EUCAST breakpoints. Following retesting of discrepant results, concordance between phenotypic results and NGD genetic predictions was 99.69%. Further investigation of 22 isolate genomes associated with persistent discrepancies revealed a range of reasons in 12 cases, but no cause could be found for the remainder. Genetic predictions generated by the NGD tool were compared with predictions generated by an independent research-based informatics approach, which demonstrated an overall concordance between the two methods of 99.97%. CONCLUSIONS: We conclude that the NGD system provides rapid and accurate prediction of the antibiotic susceptibility of MRSA.

Pilot Evaluation of a Fully Automated Bioinformatics System for Analysis of Methicillin-Resistant Staphylococcus aureus Genomes and Detection of Outbreaks.

Genomic surveillance that combines bacterial sequencing and epidemiological information will become the gold standard for outbreak detection, but its clinical translation is hampered by the lack of automated interpretation tools. We performed a prospective pilot study to evaluate the analysis of methicillin-resistant Staphylococcus aureus (MRSA) genomes using the Next Gen Diagnostics (NGD) automated bioinformatics system. Seventeen unselected MRSA-positive patients were identified in a clinical microbiology laboratory in England over a period of 2 weeks in 2018, and 1 MRSA isolate per case was sequenced on the Illumina MiniSeq instrument. The NGD system automatically activated after sequencing and processed fastq folders to determine species, multilocus sequence type, the presence of a mec gene, antibiotic susceptibility predictions, and genetic relatedness based on mapping to a reference MRSA genome and detection of pairwise core genome single-nucleotide polymorphisms. The NGD system required 90 s per sample to automatically analyze data from each run, the results of which were automatically displayed. The same data were independently analyzed using a research-based approach. There was full concordance between the two analysis methods regarding species (S. aureus), detection of mecA, sequence type assignment, and detection of genetic determinants of resistance. Both analysis methods identified two MRSA clusters based on relatedness, one of which contained 3 cases that were involved in an outbreak linked to a clinic and ward associated with diabetic patient care. We conclude that, in this pilot study, the NGD system provided rapid and accurate data that could support infection control practices.

Bacterial culture detection and identification in blood agar plates with an optoelectronic nose.

Clinical microbiology automation is currently limited by the lack of an in-plate culture identification system. Using an inexpensive, printed, disposable colorimetric sensor array (CSA) responsive to the volatiles emitted into plate headspace by microorganisms during growth, we report here that not only the presence but the species of bacteria growing in plate was identified before colonies are visible. In 1894 trials, 15 pathogenic bacterial species cultured on blood agar were identified with 91.0% sensitivity and 99.4% specificity within 3 hours of detection. The results indicate CSAs integrated into Petri dish lids present a novel paradigm to speciate microorganisms, well-suited to integration into automated plate handling systems.

Large-Scale Evaluation of a Rapid Fully Automated Analysis Platform to Detect and Refute Outbreaks Based on MRSA Genome Comparisons.

Genomic epidemiology of methicillin-resistant Staphylococcus aureus (MRSA) could transform outbreak investigations, but its clinical introduction is hampered by the lack of automated data analysis tools to rapidly and accurately define transmission based on sequence relatedness. We aimed to evaluate a fully automated bioinformatics system for MRSA genome analysis versus a bespoke researcher-led manual informatics pipeline. We analyzed 781 MRSA genomes from 777 consecutive patients identified over a 9-month period in a clinical microbiology laboratory in the United Kingdom. Outputs were bacterial species identification, detection of mec genes, assignment to sequence types (STs), identification of pairwise relatedness using a definition of ≤25 single nucleotide polymorphisms (SNPs) apart, and use of genetic relatedness to identify clusters. There was full concordance between the two analysis methods for species identification, detection of mec genes, and ST assignment. A total of 3,311 isolate pairs ≤25 SNPs apart were identified by at least one method. These had a median (range) SNP difference between the two methods of 1.2 SNPs (0 to 22 SNPs), with most isolate pairs (87%) varying by ≤2 SNPs. This similarity increased when the research pipeline was modified to use a clonal-complex-specific reference (median 0 SNP difference, 91% varying by ≤2 SNPs). Both pipelines clustered 338 isolates/334 patients into 66 unique clusters based on genetic relatedness. We conclude that the automated transmission detection tool worked at least as well as a researcher-led manual analysis and indicates how such tools could support the rapid use of MRSA genomic epidemiology in infection control practice. IMPORTANCE It has been clearly established that genome sequencing of MRSA improves the accuracy of health care-associated outbreak investigations, including the confirmation and exclusion of outbreaks and identification of patients involved. This could lead to more targeted infection control actions but its use in clinical practice is prevented by several barriers, one of which is the availability of genome analysis tools that do not depend on specialist knowledge to analyze or interpret the results. We evaluated a prototype of a fully automated bioinformatics system for MRSA genome analysis versus a bespoke researcher-led manual informatics pipeline, using genomes from 777 patients over a period of 9 months. The performance was at least equivalent to the researcher-led manual genomic analysis. This indicates the feasibility of automated analysis and represents one more step toward the routine use of pathogen sequencing in infection prevention and control practice.

The combined rapid detection and species-level identification of yeasts in simulated blood culture using a colorimetric sensor array.

BACKGROUND: A colorimetric sensor array (CSA) has been demonstrated to rapidly detect and identify bacteria growing in blood cultures by obtaining a species-specific "fingerprint" of the volatile organic compounds (VOCs) produced during growth. This capability has been demonstrated in prokaryotes, but has not been reported for eukaryotic cells growing in culture. The purpose of this study was to explore if a disposable CSA could differentially identify 7 species of pathogenic yeasts growing in blood culture. METHODS: Culture trials of whole blood inoculated with a panel of clinically important pathogenic yeasts at four different microorganism loads were performed. Cultures were done in both standard BacT/Alert and CSA-embedded bottles, after adding 10 mL of spiked blood to each bottle. Color changes in the CSA were captured as images by an optical scanner at defined time intervals. The captured images were analyzed to identify the yeast species. Time to detection by the CSA was compared to that in the BacT/Alert system. RESULTS: One hundred sixty-two yeast culture trials were performed, including strains of several species of Candida (Ca. albicans, Ca. glabrata, Ca. parapsilosis, and Ca. tropicalis), Clavispora (synonym Candida) lusitaniae, Pichia kudriavzevii (synonym Candida krusei) and Cryptococcus neoformans, at loads of 8.2 × 105, 8.3 × 103, 8.5 × 101, and 1.7 CFU/mL. In addition, 8 negative trials (no yeast) were conducted. All negative trials were correctly identified as negative, and all positive trials were detected. Colorimetric responses were species-specific and did not vary by inoculum load over the 500000-fold range of loads tested, allowing for accurate species-level identification. The mean sensitivity for species-level identification by CSA was 74% at detection, and increased with time, reaching almost 95% at 4 hours after detection. At an inoculum load of 1.7 CFU/mL, mean time to detection with the CSA was 6.8 hours (17%) less than with the BacT/Alert platform. CONCLUSION: The CSA combined rapid detection of pathogenic yeasts in blood culture with accurate species-level identification.

Rapid Diagnosis of Tuberculosis from Analysis of Urine Volatile Organic Compounds.

The World Health Organization has called for simple, sensitive, and non-sputum diagnostics for tuberculosis. We report development of a urine tuberculosis test using a colorimetric sensor array (CSA). The sensor comprised of 73 different indicators captures high-dimensional, spatiotemporal signatures of volatile chemicals emitted by human urine samples. The sensor responses to 63 urine samples collected from 22 tuberculosis cases and 41 symptomatic controls were measured under five different urine test conditions. Basified testing condition yielded the best accuracy with 85.5% sensitivity and 79.5% specificity. The CSA urine assay offers desired features needed for tuberculosis diagnosis in endemic settings.

The use of colorimetric sensor arrays to discriminate between pathogenic bacteria.

A colorimetric sensor array is a high-dimensional chemical sensor that is cheap, compact, disposable, robust, and easy to operate, making it a good candidate technology to detect pathogenic bacteria, especially potential bioterrorism agents like Yersinia pestis and Bacillus anthracis which feature on the Center for Disease Control and Prevention's list of potential biothreats. Here, a colorimetric sensor array was used to continuously monitor the volatile metabolites released by bacteria in solid media culture in an Advisory Committee on Dangerous Pathogen Containment Level 3 laboratory. At inoculum concentrations as low as 8 colony-forming units per plate, 4 different bacterial species were identified with 100% accuracy using logistic regression to classify the kinetic profile of sensor responses to culture headspace gas. The sensor array was able to further discriminate between different strains of the same species, including 5 strains of Yersinia pestis and Bacillus anthracis. These preliminary results suggest that disposable colorimetric sensor arrays can be an effective, low-cost tool to identify pathogenic bacteria.

Evolving a neural olfactorimotor system in virtual and real olfactory environments.

To provide a platform to enable the study of simulated olfactory circuitry in context, we have integrated a simulated neural olfactorimotor system with a virtual world which simulates both computational fluid dynamics as well as a robotic agent capable of exploring the simulated plumes. A number of the elements which we developed for this purpose have not, to our knowledge, been previously assembled into an integrated system, including: control of a simulated agent by a neural olfactorimotor system; continuous interaction between the simulated robot and the virtual plume; the inclusion of multiple distinct odorant plumes and background odor; the systematic use of artificial evolution driven by olfactorimotor performance (e.g., time to locate a plume source) to specify parameter values; the incorporation of the realities of an imperfect physical robot using a hybrid model where a physical robot encounters a simulated plume. We close by describing ongoing work toward engineering a high dimensional, reversible, low power electronic olfactory sensor which will allow olfactorimotor neural circuitry evolved in the virtual world to control an autonomous olfactory robot in the physical world. The platform described here is intended to better test theories of olfactory circuit function, as well as provide robust odor source localization in realistic environments.

Biomimetic chemical sensors using nanoelectronic readout of olfactory receptor proteins.

We have designed and implemented a practical nanoelectronic interface to G-protein coupled receptors (GPCRs), a large family of membrane proteins whose roles in the detection of molecules outside eukaryotic cells make them important pharmaceutical targets. Specifically, we have coupled olfactory receptor proteins (ORs) with carbon nanotube transistors. The resulting devices transduce signals associated with odorant binding to ORs in the gas phase under ambient conditions and show responses that are in excellent agreement with results from established assays for OR-ligand binding. The work represents significant progress on a path toward a bioelectronic nose that can be directly compared to biological olfactory systems as well as a general method for the study of GPCR function in multiple domains using electronic readout.

Recoding patterns of sensory input: higher-order features and the function of nonlinear dendritic trees.

Here analytical and simulation results are presented characterizing the recoding arising when overlapping patterns of sensor input impinge on an array of model neurons with branched thresholded dendritic trees. Thus, the neural units employed are intended to capture the integrative behavior of pyramidal cells that sustain isolated Na(+) or NMDA spikes in their branches. Given a defined set of sensor vectors, equations were derived for the probability of firing of both branches and neurons and for the expected overlap between the neural firing patterns triggered by two afferent patterns of given overlap. Thus, both the sparseness of the neural representation and the orthogonalization of overlapping vectors were computed. Simulations were then performed with an array of 1000 neurons comprising 30,000 branches to verify the analytical results and confirm their applicability to systems (which include any practicable artificial system) in which the combinatorically possible branches and neurons are severely subsampled. A means of readout and a measure of discrimination performance were provided so that the accuracy of discrimination among overlapping sensor vectors could be optimized as a function of neuron structure parameters. Good performance required both orthogonalization of the afferent patterns, so that discrimination was accurate and free of interference, and maintenance of a minimum level of neural activity, so that some neurons fired in response to each sensor pattern. It is shown that the discrimination performance achieved by arrays of neurons with branched dendritic trees could not be reached with single-compartment units, regardless of how many of the latter are used. The analytical results furnish a benchmark against which to measure further enhancements in the performance of subsequent simulated systems incorporating local neural mechanisms which, while often less amenable to closed-form analysis, are ubiquitous in biological neural circuitry.

A model of thalamocortical relay cells.

It is well established that the main intrinsic electrophysiological properties of thalamocortical relay cells, production of a low threshold burst upon release from hyperpolarized potential and production of a train of single spikes following stimulation from depolarized potentials, can be readily modelled using a single compartment. There is, however, another less well explored intrinsic electrophysiological characteristic of relay cells for which models have not yet accounted: at somatic potentials near spike threshold, relay cells produce a fast ragged high threshold oscillation in somatic voltage. Optical [Ca(2+)] imaging and pharmacological tests indicate that this oscillation correlates with a high threshold Ca(2+) current in the dendrites. Here we present the development of a new compartment model of the thalamic relay cell guided by the simultaneous constraints that it must produce the familiar regular spiking relay mode and low threshold rebound bursts which characterize these cells, as well as the less-studied fast oscillation occurring at near-threshold somatic potentials. We arrive at a model cell which is capable of the production of isolated high threshold Ca(2+) spikes in distal branch segments, driven by a rapidly inactivating intermediate threshold Ca(2+) channel. Further, the model produces the low threshold spike behaviour of the relay cell without requiring high T-current density in the distal dendritic segments. The results thus support a new picture of the dendritic tree of relay cells which may have implications for the manner in which thalamic relay cells integrate descending input from the cortex.