Machine learning-based predictive modeling to identify genotypic traits associated with Salmonella enterica disease endpoints in isolates from ground chicken

Using GWAS and Machine Learning to Identify and Predict Genetic Variants Associated with Foodborne Bacteria Phenotypic Traits

One of the main challenges in food microbiology is to prevent the risk of outbreaks by avoiding the distribution of food contaminated by bacteria. This requires constant monitoring of the circulating strains throughout the food production chain. Bacterial genomes contain signatures of natural evolution and adaptive markers that can be exploited to better understand the behavior of pathogen in the food industry. The monitoring of foodborne strains can therefore be facilitated by the use of these genomic markers capable of rapidly providing essential information on isolated strains, such as the source of contamination, risk of illness, potential for biofilm formation, and tolerance or resistance to biocides. The increasing availability of large genome datasets is enhancing the understanding of the genetic basis of complex traits such as host adaptation, virulence, and persistence. Genome-wide association studies have shown very promising results in the discovery of genomic markers that can be integrated into rapid detection tools. In addition, machine learning has successfully predicted phenotypes and classified important traits. Genome-wide association and machine learning tools have therefore the potential to support decision-making circuits intending at reducing the burden of foodborne diseases. The aim of this chapter review is to provide knowledge on the use of these two methods in food microbiology and to recommend their use in the field.

Research gaps and priorities for quantitative microbial risk assessment (QMRA)

The coronavirus disease 2019 pandemic highlighted the need for more rapid and routine application of modeling approaches such as quantitative microbial risk assessment (QMRA) for protecting public health. QMRA is a transdisciplinary science dedicated to understanding, predicting, and mitigating infectious disease risks. To better equip QMRA researchers to inform policy and public health management, an Advances in Research for QMRA workshop was held to synthesize a path forward for QMRA research. We summarize insights from 41 QMRA researchers and experts to clarify the role of QMRA in risk analysis by (1) identifying key research needs, (2) highlighting emerging applications of QMRA; and (3) describing data needs and key scientific efforts to improve the science of QMRA. Key identified research priorities included using molecular tools in QMRA, advancing dose-response methodology, addressing needed exposure assessments, harmonizing environmental monitoring for QMRA, unifying a divide between disease transmission and QMRA models, calibrating and/or validating QMRA models, modeling co-exposures and mixtures, and standardizing practices for incorporating variability and uncertainty throughout the source-to-outcome continuum. Cross-cutting needs identified were to: develop a community of research and practice, integrate QMRA with other scientific approaches, increase QMRA translation and impacts, build communication strategies, and encourage sustainable funding mechanisms. Ultimately, a vision for advancing the science of QMRA is outlined for informing national to global health assessments, controls, and policies.

Lack of correlation between growth, stress, and virulence phenotypes in strains of Salmonella enterica serovar Enteritidis, S. Typhimurium DT104, S. 4,12, b:- and S. Liverpool

Strains of Salmonella Enteritidis (SEnt, n = 10) and S. Typhimurium (STm, n = 11), representing clones with high impact on human health, and strains of S. 4,12: b:- (S412B n = 11) and S. Liverpool (SLiv, n = 4), representing clones with minor impact on human health were characterized for 16 growth, stress, and virulence phenotypes to investigate whether systematic differences exist in their performance in these phenotypes and whether there was correlation between performance in different phenotypes. The term serotype was not found to be predictive of a certain type of performance in any phenotype, and surprisingly, on average, strains of SEnt and STm were not significantly better in adhering to and invading cultured intestinal cells than the less pathogenic types. Forest analysis identified desiccation tolerance and the ability to grow at 42°C with high salt as the characters that separated serovars with low human health impact (S412B/SLiv) from serovars with high human health impact (SEnt/STm). The study showed that variation in phenotypes was high even within serovars and correlation between phenotypes was low, i.e. the way that a strain performed phenotypically in one of the tested conditions had a low predictive value for the performance of the strain in other conditions.

A machine learning approach to identifying Salmonella stress response genes in isolates from poultry processing

We explored the potential of machine learning to identify significant genes associated with Salmonella stress response during poultry processing using whole genome sequencing (WGS) data. The Salmonella isolates (n = 177) used in this study were obtained from various chicken sources (skin before chiller, chicken carcass before chiller, frozen chicken, and post-chill chicken carcass). Six machine learning algorithms (random forest, neural network, cost-sensitive learning, logit boost, and support vector machine linear and radial kernels) were trained on Salmonella WGS data, and model fit was assessed using standard evaluation metrics such as the area under the receiver operating characteristic (AUROC) curve and confusion matrix statistics. All models achieved high performances based on the AUROC metric, with logit boost showing the best performance with an AUROC score of 0.904, sensitivity of 0.889, and specificity of 0.920. The significant genes identified included ybtX, which encodes a Yersiniabactin-associated zinc transporter, and the transferase-encoding genes yccK and thiS. Additionally, genes coding for cold (cspA, cspD, and cspE) and heat shock (rpoH and rpoE) responses were identified. Other significant genes included those involved in lipopolysaccharide biosynthesis (irp1, waaD, rfc, and rfbX), DNA repair and replication (traI), biofilm formation (ccdA and fyuA), and cellular metabolism (irtA).

Comparison of Source Attribution Methodologies for Human Campylobacteriosis

Campylobacter spp. are the most common cause of bacterial gastrointestinal infection in humans both in Denmark and worldwide. Studies have found microbial subtyping to be a powerful tool for source attribution, but comparisons of different methodologies are limited. In this study, we compare three source attribution approaches (Machine Learning, Network Analysis, and Bayesian modeling) using three types of whole genome sequences (WGS) data inputs (cgMLST, 5-Mers and 7-Mers). We predicted and compared the sources of human campylobacteriosis cases in Denmark. Using 7mer as an input feature provided the best model performance. The network analysis algorithm had a CSC value of 78.99% and an F1-score value of 67%, while the machine-learning algorithm showed the highest accuracy (98%). The models attributed between 965 and all of the 1224 human cases to a source (network applying 5mer and machine learning applying 7mer, respectively). Chicken from Denmark was the primary source of human campylobacteriosis with an average percentage probability of attribution of 45.8% to 65.4%, representing Bayesian with 7mer and machine learning with cgMLST, respectively. Our results indicate that the different source attribution methodologies based on WGS have great potential for the surveillance and source tracking of Campylobacter. The results of such models may support decision makers to prioritize and target interventions.

Development of prediction software to describe total mesophilic bacteria in spinach using a machine learning-based regression approach

The purpose of this study was to create a tool for predicting the growth of total mesophilic bacteria in spinach using machine learning-based regression models such as support vector regression, decision tree regression, and Gaussian process regression. The performance of these models was compared to traditionally used models (modified Gompertz, Baranyi, and Huang models) using statistical indices like the coefficient of determination (R2) and root mean square error (RMSE). The results showed that the machine learning-based regression models provided more accurate predictions with an R2 of at least 0.960 and an RMSE of at most 0.154, indicating that they can be used as an alternative to traditional approaches for predictive total mesophilic. Therefore, the developed software in this work has a significant potential to be used as an alternative simulation method to traditionally used approach in the predictive food microbiology field.

Development of a novel machine learning-based weighted modeling approach to incorporate Salmonella enterica heterogeneity on a genetic scale in a dose-response modeling framework

Estimating microbial dose-response is an important aspect of a food safety risk assessment. In recent years, there has been considerable interest to advance these models with potential incorporation of gene expression data. The aim of this study was to develop a novel machine learning model that considers the weights of expression of Salmonella genes that could be associated with illness, given exposure, in hosts. Here, an elastic net-based weighted Poisson regression method was proposed to identify Salmonella enterica genes that could be significantly associated with the illness response, irrespective of serovar. The best-fit elastic net model was obtained by 10-fold cross-validation. The best-fit elastic net model identified 33 gene expression-dose interaction terms that added to the predictability of the model. Of these, nine genes associated with Salmonella metabolism and virulence were found to be significant by the best-fit Poisson regression model (p < 0.05). This method could improve or redefine dose-response relationships for illness from relative proportions of significant genes from a microbial genetic dataset, which would help in refining endpoint and risk estimations.

Machine learning-based typing of Salmonella enterica O-serogroups by the Fourier-Transform Infrared (FTIR) Spectroscopy-based IR Biotyper system

BACKGROUND

Salmonella enterica is among the major burdens for public health at global level. Typing of salmonellae below the species level is fundamental for different purposes, but traditional methods are expensive, technically demanding, and time-consuming, and therefore limited to reference centers. Fourier transform infrared (FTIR) spectroscopy is an alternative method for bacterial typing, successfully applied for classification at different infra-species levels.

AIM

This study aimed to address the challenge of subtyping Salmonella enterica at O-serogroup level by using FTIR spectroscopy. We applied machine learning to develop a novel approach for S. enterica typing, using the FTIR-based IR Biotyper® system (IRBT; Bruker Daltonics GmbH & Co. KG, Germany). We investigated a multicentric collection of isolates, and we compared the novel approach with classical serotyping-based and molecular methods.

METHODS

A total of 958 well characterized Salmonella isolates (25 serogroups, 138 serovars), collected in 11 different centers (in Europe and Japan), from clinical, environmental and food samples were included in this study and analyzed by IRBT. Infrared absorption spectra were acquired from water-ethanol bacterial suspensions, from culture isolates grown on seven different agar media. In the first part of the study, the discriminatory potential of the IRBT system was evaluated by comparison with reference typing method/s. In the second part of the study, the artificial intelligence capabilities of the IRBT software were applied to develop a classifier for Salmonella isolates at serogroup level. Different machine learning algorithms were investigated (artificial neural networks and support vector machine). A subset of 88 pre-characterized isolates (corresponding to 25 serogroups and 53 serovars) were included in the training set. The remaining 870 samples were used as validation set. The classifiers were evaluated in terms of accuracy, error rate and failed classification rate.

RESULTS

The classifier that provided the highest accuracy in the cross-validation was selected to be tested with four external testing sets. Considering all the testing sites, accuracy ranged from 97.0% to 99.2% for non-selective media, and from 94.7% to 96.4% for selective media.

CONCLUSIONS

The IRBT system proved to be a very promising, user-friendly, and cost-effective tool for Salmonella typing at serogroup level. The application of machine learning algorithms proved to enable a novel approach for typing, which relies on automated analysis and result interpretation, and it is therefore free of potential human biases. The system demonstrated a high robustness and adaptability to routine workflows, without the need of highly trained personnel, and proving to be suitable to be applied with isolates grown on different agar media, both selective and unselective. Further tests with currently circulating clinical, food and environmental isolates would be necessary before implementing it as a potentially stand-alone standard method for routine use.

Mathematical Models for Typhoid Disease Transmission: A Systematic Literature Review

Explaining all published articles on the typhoid disease transmission model was carried out. It has been conducted to understand how Salmonella is transmitted among humans and vectors with variation interventions to control the spread of the typhoid disease. Specific objectives were to (1) identify the model developed, (2) describe the studies, and (3) identify the interventions of the model. It systemically searched and reviewed Dimension, Scopus, and ScienceDirect databases from 2013 through to 2022 for articles that studied the spread of typhoid fever through a compartmental mathematical model. This study obtained 111 unique articles from three databases, resulting in 23 articles corresponding to the created terms. All the articles were elaborated on to identify their identities for more explanation. Various interventions were considered in the model of each article, are identified, and then summarized to find out the opportunities for model development in future works. The whole article’s content was identified and outlined regarding how mathematics plays a role in model analysis and study of typhoid disease spread with various interventions. The study of mathematical modeling for typhoid disease transmission can be developed on analysis and creating the model with direct and indirect interventions to the human population for further work.

A Machine Learning Model for Food Source Attribution of Listeria monocytogenes

Despite its low morbidity, listeriosis has a high mortality rate due to the severity of its clinical manifestations. The source of human listeriosis is often unclear. In this study, we investigate the ability of machine learning to predict the food source from which clinical Listeria monocytogenes isolates originated. Four machine learning classification algorithms were trained on core genome multilocus sequence typing data of 1212 L. monocytogenes isolates from various food sources. The average accuracies of random forest, support vector machine radial kernel, stochastic gradient boosting, and logit boost were found to be 0.72, 0.61, 0.7, and 0.73, respectively. Logit boost showed the best performance and was used in model testing on 154 L. monocytogenes clinical isolates. The model attributed 17.5 % of human clinical cases to dairy, 32.5% to fruits, 14.3% to leafy greens, 9.7% to meat, 4.6% to poultry, and 18.8% to vegetables. The final model also provided us with genetic features that were predictive of specific sources. Thus, this combination of genomic data and machine learning-based models can greatly enhance our ability to track L. monocytogenes from different food sources.

Source Attribution of Human Campylobacteriosis Using Whole-Genome Sequencing Data and Network Analysis

Campylobacter spp. are a leading and increasing cause of gastrointestinal infections worldwide. Source attribution, which apportions human infection cases to different animal species and food reservoirs, has been instrumental in control- and evidence-based intervention efforts. The rapid increase in whole-genome sequencing data provides an opportunity for higher-resolution source attribution models. Important challenges, including the high dimension and complex structure of WGS data, have inspired concerted research efforts to develop new models. We propose network analysis models as an accurate, high-resolution source attribution approach for the sources of human campylobacteriosis. A weighted network analysis approach was used in this study for source attribution comparing different WGS data inputs. The compared model inputs consisted of cgMLST and wgMLST distance matrices from 717 human and 717 animal isolates from cattle, chickens, dogs, ducks, pigs and turkeys. SNP distance matrices from 720 human and 720 animal isolates were also used. The data were collected from 2015 to 2017 in Denmark, with the animal sources consisting of domestic and imports from 7 European countries. Clusters consisted of network nodes representing respective genomes and links representing distances between genomes. Based on the results, animal sources were the main driving factor for cluster formation, followed by type of species and sampling year. The coherence source clustering (CSC) values based on animal sources were 78%, 81% and 78% for cgMLST, wgMLST and SNP, respectively. The CSC values based on Campylobacter species were 78%, 79% and 69% for cgMLST, wgMLST and SNP, respectively. Including human isolates in the network resulted in 88%, 77% and 88% of the total human isolates being clustered with the different animal sources for cgMLST, wgMLST and SNP, respectively. Between 12% and 23% of human isolates were not attributed to any animal source. Most of the human genomes were attributed to chickens from Denmark, with an average attribution percentage of 52.8%, 52.2% and 51.2% for cgMLST, wgMLST and SNP distance matrices respectively, while ducks from Denmark showed the least attribution of 0% for all three distance matrices. The best-performing model was the one using wgMLST distance matrix as input data, which had a CSC value of 81%. Results from our study show that the weighted network-based approach for source attribution is reliable and can be used as an alternative method for source attribution considering the high performance of the model. The model is also robust across the different Campylobacter species, animal sources and WGS data types used as input.