Nontyphoidal Salmonella species are the leading bacterial cause of foodborne disease in the United States. Whole-genome sequences and paired antimicrobial susceptibility data are available for Salmonella strains because of surveillance efforts from public health agencies. In this study, a collection of 5,278 nontyphoidal Salmonella genomes, collected over 15 years in the United States, was used to generate extreme gradient boosting (XGBoost)-based machine learning models for predicting MICs for 15 antibiotics. The MIC prediction models had an overall average accuracy of 95% within ±1 2-fold dilution step (confidence interval, 95% to 95%), an average very major error rate of 2.7% (confidence interval, 2.4% to 3.0%), and an average major error rate of 0.1% (confidence interval, 0.1% to 0.2%). The model predicted MICs with no a priori information about the underlying gene content or resistance phenotypes of the strains. By selecting diverse genomes for the training sets, we show that highly accurate MIC prediction models can be generated with less than 500 genomes. We also show that our approach for predicting MICs is stable over time, despite annual fluctuations in antimicrobial resistance gene content in the sampled genomes. Finally, using feature selection, we explore the important genomic regions identified by the models for predicting MICs. To date, this is one of the largest MIC modeling studies to be published. Our strategy for developing whole-genome sequence-based models for surveillance and clinical diagnostics can be readily applied to other important human pathogens.

Using machine learning to predict antimicrobial MICs and associated genomic features for nontyphoidal Salmonella

Background: Tracking the spread of antimicrobial-resistant Neisseria gonorrhoeae is a major priority for national surveillance programmes. / Objectives: We investigate whether WGS and simultaneous analysis of multiple resistance determinants can be used to predict antimicrobial susceptibilities to the level of MICs in N. gonorrhoeae. / Methods: WGS was used to identify previously reported potential resistance determinants in 681 N. gonorrhoeae isolates, from England, the USA and Canada, with phenotypes for cefixime, penicillin, azithromycin, ciprofloxacin and tetracycline determined as part of national surveillance programmes. Multivariate linear regression models were used to identify genetic predictors of MIC. Model performance was assessed using leave-one-out cross-validation. / Results: Overall 1785/3380 (53%) MIC values were predicted to the nearest doubling dilution and 3147 (93%) within ±1 doubling dilution and 3314 (98%) within ±2 doubling dilutions. MIC prediction performance was similar across the five antimicrobials tested. Prediction models included the majority of previously reported resistance determinants. Applying EUCAST breakpoints to MIC predictions, the overall very major error (VME; phenotypically resistant, WGS-prediction susceptible) rate was 21/1577 (1.3%, 95% CI 0.8%–2.0%) and the major error (ME; phenotypically susceptible, WGS-prediction resistant) rate was 20/1186 (1.7%, 1.0%–2.6%). VME rates met regulatory thresholds for all antimicrobials except cefixime and ME rates for all antimicrobials except tetracycline. Country of testing was a strongly significant predictor of MIC for all five antimicrobials. / Conclusions: We demonstrate a WGS-based MIC prediction approach that allows reliable MIC prediction for five gonorrhoea antimicrobials. Our approach should allow reasonably precise prediction of MICs for a range of bacterial species.

WGS to predict antibiotic MICs for Neisseria gonorrhoeae.

Rapid advances in DNA sequencing technology ("next-generation sequencing") have inspired optimism about the potential of human genomics for "precision medicine." Meanwhile, pathogen genomics is already delivering "precision public health" through more effective investigations of outbreaks of foodborne illnesses, better-targeted tuberculosis control, and more timely and granular influenza surveillance to inform the selection of vaccine strains. In this article, we describe how public health agencies have been adopting pathogen genomics to improve their effectiveness in almost all domains of infectious disease. This momentum is likely to continue, given the ongoing development in sequencing and sequencing-related technologies.

https://www.nejm.org/doi/pdf/10.1056/NEJMsr1813907?listPDF=true

Pathogen Genomics in Public Health

Summary Background Invasive pneumococcal disease remains an important health priority owing to increasing disease incidence caused by pneumococci expressing non-vaccine serotypes. We previously defined 621 Global Pneumococcal Sequence Clusters (GPSCs) by analysing 20 027 pneumococcal isolates collected worldwide and from previously published genomic data. In this study, we aimed to investigate the pneumococcal lineages behind the predominant serotypes, the mechanism of serotype replacement in disease, as well as the major pneumococcal lineages contributing to invasive pneumococcal disease in the post-vaccine era and their antibiotic resistant traits. Methods We whole-genome sequenced 3233 invasive pneumococcal disease isolates from laboratory-based surveillance programmes in Hong Kong (n=78), Israel (n=701), Malawi (n=226), South Africa (n=1351), The Gambia (n=203), and the USA (n=674). The genomes represented pneumococci from before and after pneumococcal conjugate vaccine (PCV) introductions and were from children younger than 3 years. We identified predominant serotypes by prevalence and their major contributing lineages in each country, and assessed any serotype replacement by comparing the incidence rate between the pre-PCV and PCV periods for Israel, South Africa, and the USA. We defined the status of a lineage as vaccine-type GPSC (≥50% 13-valent PCV [PCV13] serotypes) or non-vaccine-type GPSC (>50% non-PCV13 serotypes) on the basis of its initial serotype composition detected in the earliest vaccine period to measure their individual contribution toward serotype replacement in each country. Major pneumococcal lineages in the PCV period were identified by pooled incidence rate using a random effects model. Findings The five most prevalent serotypes in the PCV13 period varied between countries, with only serotypes 5, 12F, 15B/C, 19A, 33F, and 35B/D common to two or more countries. The five most prevalent serotypes in the PCV13 period varied between countries, with only serotypes 5, 12F, 15B/C, 19A, 33F, and 35B/D common to two or more countries. These serotypes were associated with more than one lineage, except for serotype 5 (GPSC8). Serotype replacement was mainly mediated by expansion of non-vaccine serotypes within vaccine-type GPSCs and, to a lesser extent, by increases in non-vaccine-type GPSCs. A globally spreading lineage, GPSC3, expressing invasive serotypes 8 in South Africa and 33F in the USA and Israel, was the most common lineage causing non-vaccine serotype invasive pneumococcal disease in the PCV13 period. We observed that same prevalent non-vaccine serotypes could be associated with distinctive lineages in different countries, which exhibited dissimilar antibiotic resistance profiles. In non-vaccine serotype isolates, we detected significant increases in the prevalence of resistance to penicillin (52 [21%] of 249 vs 169 [29%] of 575, p=0·0016) and erythromycin (three [1%] of 249 vs 65 [11%] of 575, p=0·0031) in the PCV13 period compared with the pre-PCV period. Interpretation Globally spreading lineages expressing invasive serotypes have an important role in serotype replacement, and emerging non-vaccine serotypes associated with different pneumococcal lineages in different countries might be explained by local antibiotic-selective pressures. Continued genomic surveillance of the dynamics of the pneumococcal population with increased geographical representation in the post-vaccine period will generate further knowledge for optimising future vaccine design. Funding Bill & Melinda Gates Foundation, Wellcome Sanger Institute, and the US Centers for Disease Control.

Pneumococcal lineages associated with serotype replacement and antibiotic resistance in childhood invasive pneumococcal disease in the post-PCV13 era: an international whole-genome sequencing study.

Anaerobic bacteria have pivotal roles in the microbiota of humans and they are significant infectious agents involved in many pathological processes, both in immunocompetent and immunocompromised individuals. Their isolation, cultivation and correct identification differs significantly from the workup of aerobic species, although the use of new technologies (e.g., matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, whole genome sequencing) changed anaerobic diagnostics dramatically. In the past, antimicrobial susceptibility of these microorganisms showed predictable patterns and empirical therapy could be safely administered but recently a steady and clear increase in the resistance for several important drugs (β-lactams, clindamycin) has been observed worldwide. For this reason, antimicrobial susceptibility testing of anaerobic isolates for surveillance purposes or otherwise is of paramount importance but the availability of these testing methods is usually limited. In this present review, our aim was to give an overview of the methods currently available for the identification (using phenotypic characteristics, biochemical testing, gas-liquid chromatography, MALDI-TOF MS and WGS) and antimicrobial susceptibility testing (agar dilution, broth microdilution, disk diffusion, gradient tests, automated systems, phenotypic and molecular resistance detection techniques) of anaerobes, when should these methods be used and what are the recent developments in resistance patterns of anaerobic bacteria.

https://www.mdpi.com/2079-6382/6/4/25/pdf

Identification and Antimicrobial Susceptibility Testing of Anaerobic Bacteria: Rubik's Cube of Clinical Microbiology?

Using whole genome sequencing to identify resistance determinants and predict antimicrobial resistance phenotypes for year 2015 invasive pneumococcal disease isolates recovered in the United States

Y. Li

Papers

Using whole genome sequencing to identify resistance determinants and predict antimicrobial resistance phenotypes for year 2015 invasive pneumococcal disease isolates recovered in the United States