Genomic Epidemiology with Mixed Samples

Abstract

Genomic epidemiology is an established tool for investigation of outbreaks of infectious diseases and wider public health applications. It traces transmission of pathogens based on whole-genome sequencing of colony picks from culture plates enriching the target organism(s). In this article, we introduce the mGEMS pipeline for performing genomic epidemiology directly with plate sweeps representing mixed samples of the target pathogen in a culture plate, skipping the colony pick step entirely. By requiring only a single culturing and library preparation step per analyzed sample, we address several key issues in the current approach relating to its cost, practical application and sensitivity. Our pipeline significantly improves upon the state-of-the-art in analysing mixed short-read sequencing data from bacteria, reaching accuracy levels in downstream analyses closely resembling colony pick sequencing data that allow reliable SNP calling and subsequent phylogenetic analyses. The fundamental novel parts enabling these analyses are the mGEMS read binner for probabilistic assignments of sequencing reads and the high-throughput exact pseudoaligner Themisto. In conjunction with recent advances in probabilistic modelling of mixed bacterial samples and genome assembly techniques, these tools form the mGEMS pipeline. We demonstrate the effectiveness of our approach using closely related samples in a nosocomial setting for the three major pathogens Enterococcus faecalis , Escherichia coli and Staphylococcus aureus . Our results lend firm support to more widespread consideration of genomic epidemiology with mixed infection samples.

Pneumococcal within-host diversity during colonisation, transmission and treatment

“…The PopPUNK algorithm was used to assign each of these genomes to their respective Global Pneumococcal Sequencing Cluster (89). The mSWEEP and mGEMS pipelines were then run using the fastq files for each deep sequencing sample with the exact commands used given in the Rmarkdown provided as part of the accompanying GitHub repository (87,88). To reduce the possibility of false positives lineages were only called if they were present at a frequency of at least 1%.…”

Linear-time Minimization of Wheeler DFAs

“…For example, the compressed index structure of the variation graph toolkit VG [7] is based on the WDFA of a compacted de Bruijn graph of the input data. Other applications include the colored de Bruijn graph indexes VARI [8] and Themisto [9]. Efficient WDFA minimization has applications in compressing the space of all these data structures.…”

Tailoring r-index for metagenomics

“…One important step in metagenomics is to assign each fragment to its owner, allowing to identify and quantify species. This step is called read assignment [19], and it is the basic step in most metagenomic analysis workflows such as in genomic epidemiology [25], and viral epidemiology [6].…”

“…Efficient indexing of k-mer sets has been deeply investigated and we refer the reader to the survey [27] for further reading. Pseudoaligners such as Kallisto [4], MetaKallisto [35], and Themisto [25] are based on the following pseudoalignment criterion. Given a set of references T 1 , .…”

“…, T t . This approach and its solutions using colored de Bruijn graphs [4,35,25] are motivated by the fact that the species are usually quite dissimilar, but the strains inside the species are highly similar.…”