食中毒の発生を追跡するための新しいツール

タイトル: DODGE: Automated point source bacterial outbreak detection using cumulative long term genomic surveillance.

概要: SummaryThe reliable and timely recognition of outbreaks is a key component of public health surveillance for foodborne diseases. Whole genome sequencing (WGS) offers high resolution typing of foodborne bacterial pathogens and facilitates the accurate detection of outbreaks. This detection relies on grouping WGS data into clusters at an appropriate genetic threshold, however, methods and tools for selecting and adjusting such thresholds according to the required resolution of surveillance and epidemiological context are lacking. Here we present DODGE (Dynamic Outbreak Detection for Genomic Epidemiology), an algorithm to dynamically select and compare these genetic thresholds. DODGE can analyse expanding datasets over time and clusters that are predicted to correspond to outbreaks (or investigation clusters) can be named with the established genomic nomenclature systems to facilitate integrated analysis across jurisdictions. DODGE was tested in two real-world genomic surveillance datasets of different duration, two months from Australia and nine years from the UK. In both cases only a minority of isolates were identified as investigation clusters. Two known outbreaks in the UK dataset were detected by DODGE and were recognised at an earlier timepoint than the outbreaks were reported. These findings demonstrated the potential of the DODGE approach to improve the effectiveness and timeliness of genomic surveillance for foodborne diseases and the effectiveness of the algorithm developed. Availability and implementationDODGE is freely available at https://github.com/LanLab/dodge and can easily be installed using Conda. Supplementary informationSupplementary Tables, Results, Figure 1 and Figure 2 O_FIG O_LINKSMALLFIG WIDTH=193 HEIGHT=200 SRC="FIGDIR/small/24301506v1_fig1.gif" ALT="Figure 1"> View larger version (48K): [email protected]@16dec4dorg.highwire.dtl.DTLVardef@1b4b3f6org.highwire.dtl.DTLVardef@930876_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFigure 1.C_FLOATNO The DODGE pipeline, algorithm and Australian dataset investigation clusters. A. Flowchart describing 6 stages of the DODGE algorithm. B. Example investigation cluster detection with the same 6 stages marked. Blue circles represent isolates, red numbered lines are genetic distances. At each genetic threshold isolates within the grey shaded area are the cluster being evaluated. C. High level schematic of the DODGE pipeline including the DODGE algorithm. Genetic data in the form of allele profiles (Enterobase or MGTdb) or SNPs (output by snippy) for isolates from a given temporal window (a week or month) are combined with previous time periods to generate a combined distance matrix. Distances between isolate pairs that are not in an optional input distance matrix (Blue arrow) are calculated and added. Clusters are identified using single linkage clustering from the distance matrix. These clusters are compared to existing investigation and non-investigation clusters from previous time periods (blue arrow) to identify expanded or unchanged investigation clusters. Remaining non investigation clusters are then used to identify novel investigation clusters using the DODGE algorithm detailed in B and C. Green boxes are input files, red outlined boxes are output files, blue arrows represent outputs from one time period used as inputs in the next. D. Investigation clusters identified from the Australian dataset over time. X axis is date of collection by week. Y axis is investigation cluster with MGT ST based ID. The area of circles is proportional to number of isolates in that investigation cluster in that week. Colour represents the genetic threshold used for that investigation cluster. Red outline indicates the week in which the cluster was identified as an investigation cluster by the DODGE algorithm. C_FIG