Is it the same strain? Defining genomic epidemiology thresholds tailored to individual outbreaks

BackgroundEpidemiological surveillance relies on microbial strain typing, which defines genomic relatedness among isolates to identify case clusters and their potential sources. No consensus exists on the choice of thresholds of genomic relatedness to define clusters. While a priori defined thresholds are often applied, outbreak-specific features such as pathogen mutation rate and duration of source contamination should be considered. MethodsWe developed a forward model of bacterial evolution to simulate mutation within a population diversifying at a specific mutation rate, with specific outbreak duration and sample isolation dates. Based on the resulting expected distribution of genetic distances we define a threshold beyond which isolates are considered as not part of the outbreak. We additionally embedded the model into a Markov Chain Monte Carlo inference framework to estimate, from data including sampling dates or isolates genetic variation, the most credible mutation rate or time since source contamination. FindingsA simulation study validated the model over realistic durations and mutation rates. When applied to 16 published datasets describing foodborne outbreaks, our framework consistently identified outliers. Appropriate thresholds for grouping cases were obtained for 14 outbreaks. For the remaining two outbreaks, re-estimation of the duration of outbreak lead to updated threshold values and was more likely, given our model, to result in the observed genetic distances. InterpretationWe propose an evolutionary approach to the single strain conundrum by defining the genetic threshold based on individual outbreak properties. The framework provides an informed estimation of the likelihood of a cluster given the samples epidemiological and microbiological context. This forward model, applicable to foodborne or environmental-source single point case clusters or outbreaks, will be useful for epidemiological surveillance and to guide control measures. FundingThis work was supported financially by the MedVetKlebs project, a component of European Joint Programme One Health EJP, which has received funding from the European Unions Horizon 2020 research and innovation programme under Grant Agreement No 773830. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSWe searched PubMed for studies published between database inception and April 3, 2021, with the term (threshold OR cut-off OR genetic relatedness) AND (outbreak) AND (cgMLST OR wgMLST OR SNPs) AND (microbial OR bacteria OR bacterial OR pathogen). We found 222 related articles. Most studies define a fixed SNP threshold that relate outbreak strains based on previous observations. One original study identifies outbreak clusters based on transmission events. However, it relies on strong assumptions about molecular clock and transmission processes. Added value of this studyOur study describes a new method based on a forward Wright-Fisher model to find the most credible genetic distance threshold. This method is fast and simple to use with only few assumptions, informed by outbreak duration and pathogen mutation rate. By using SNP or cgMLST pairwise distances and sample collection dates of the outbreak of interest, the algorithm provides context-based guidance to separate outbreak strains from outliers. Implications of all the available evidenceThe fast and easy method developed here enables to move away from a priori defined thresholds. Defining clusters more accurately based on the specific features of outbreaks, and the ability to estimate outbreak duration, will provide the needed precision for epidemiological surveillance and should contribute to leverage molecular epidemiology data more efficiently for the purpose of uncovering contamination sources. Data Availability StatementAll data and code used for this manuscript is available online at https://gitlab.pasteur.fr/BEBP.