Visualizing variation within Global Pneumococcal Sequence Clusters (GPSCs) and country population snapshots to contextualize pneumococcal isolates

Knowledge of pneumococcal lineages, their geographic distribution and antibiotic resistance patterns, can give insights into global pneumococcal disease. We provide interactive bioinformatic outputs to explore such topics, aiming to increase dissemination of genomic insights to the wider community, without the need for specialist training. We prepared 12 country-specific phylogenetic snapshots, and international phylogenetic snapshots of 73 common Global Pneumococcal Sequence Clusters (GPSCs) previously defined using PopPUNK, and present them in Microreact. Gene presence and absence defined using Roary, and recombination profiles derived from Gubbins are presented in Phandango for each GPSC. Temporal phylogenetic signal was assessed for each GPSC using BactDating. We provide examples of how such resources can be used. In our example use of a country-specific phylogenetic snapshot we determined that serotype 14 was observed in nine unrelated genetic backgrounds in South Africa. The international phylogenetic snapshot of GPSC9, in which most serotype 14 isolates from South Africa were observed, highlights that there were three independent sub-clusters represented by South African serotype 14 isolates. We estimated from the GPSC9-dated tree that the sub-clusters were each established in South Africa during the 1980s. We show how recombination plots allowed the identification of a 20 kb recombination spanning the capsular polysaccharide locus within GPSC97. This was consistent with a switch from serotype 6A to 19A estimated to have occured in the 1990s from the GPSC97-dated tree. Plots of gene presence/absence of resistance genes (tet, erm, cat) across the GPSC23 phylogeny were consistent with acquisition of a composite transposon. We estimated from the GPSC23-dated tree that the acquisition occurred between 1953 and 1975. Finally, we demonstrate the assignment of GPSC31 to 17 externally generated pneumococcal serotype 1 assemblies from Utah via Pathogenwatch. Most of the Utah isolates clustered within GPSC31 in a USA-specific clade with the most recent common ancestor estimated between 1958 and 1981. The resources we have provided can be used to explore to data, test hypothesis and generate new hypotheses. The accessible assignment of GPSCs allows others to contextualize their own collections beyond the data presented here.

Predicting the short-term success of human influenza virus variants with machine learning

Seasonal influenza viruses are constantly changing and produce a different set of circulating strains each season. Small genetic changes can accumulate over time and result in antigenically different viruses; this may prevent the body's immune system from recognizing those viruses. Due to rapid mutations, in particular, in the haemagglutinin (HA) gene, seasonal influenza vaccines must be updated frequently. This requires choosing strains to include in the updates to maximize the vaccines' benefits, according to estimates of which strains will be circulating in upcoming seasons. This is a challenging prediction task. In this paper, we use longitudinally sampled phylogenetic trees based on HA sequences from human influenza viruses, together with counts of epitope site polymorphisms in HA, to predict which influenza virus strains are likely to be successful. We extract small groups of taxa (subtrees) and use a suite of features of these subtrees as key inputs to the machine learning tools. Using a range of training and testing strategies, including training on H3N2 and testing on H1N1, we find that successful prediction of future expansion of small subtrees is possible from these data, with accuracies of 0.71-0.85 and a classifier 'area under the curve' 0.75-0.9.

Contaminant DNA in bacterial sequencing experiments is a major source of false genetic variability

Background

Contaminant DNA is a well-known confounding factor in molecular biology and in genomic repositories. Strikingly, analysis workflows for whole-genome sequencing (WGS) data commonly do not account for errors potentially introduced by contamination, which could lead to the wrong assessment of allele frequency both in basic and clinical research.

Results

We used a taxonomic filter to remove contaminant reads from more than 4000 bacterial samples from 20 different studies and performed a comprehensive evaluation of the extent and impact of contaminant DNA in WGS. We found that contamination is pervasive and can introduce large biases in variant analysis. We showed that these biases can result in hundreds of false positive and negative SNPs, even for samples with slight contamination. Studies investigating complex biological traits from sequencing data can be completely biased if contamination is neglected during the bioinformatic analysis, and we demonstrate that removing contaminant reads with a taxonomic classifier permits more accurate variant calling. We used both real and simulated data to evaluate and implement reliable, contamination-aware analysis pipelines.

Conclusion

As sequencing technologies consolidate as precision tools that are increasingly adopted in the research and clinical context, our results urge for the implementation of contamination-aware analysis pipelines. Taxonomic classifiers are a powerful tool to implement such pipelines.

Typing methods based on whole genome sequencing data

Whole genome sequencing (WGS) of foodborne pathogens has become an effective method for investigating the information contained in the genome sequence of bacterial pathogens. In addition, its highly discriminative power enables the comparison of genetic relatedness between bacteria even on a sub-species level. For this reason, WGS is being implemented worldwide and across sectors (human, veterinary, food, and environment) for the investigation of disease outbreaks, source attribution, and improved risk characterization models. In order to extract relevant information from the large quantity and complex data produced by WGS, a host of bioinformatics tools has been developed, allowing users to analyze and interpret sequencing data, starting from simple gene-searches to complex phylogenetic studies. Depending on the research question, the complexity of the dataset and their bioinformatics skill set, users can choose between a great variety of tools for the analysis of WGS data. In this review, we describe the relevant approaches for phylogenomic studies for outbreak studies and give an overview of selected tools for the characterization of foodborne pathogens based on WGS data. Despite the efforts of the last years, harmonization and standardization of typing tools are still urgently needed to allow for an easy comparison of data between laboratories, moving towards a one health worldwide surveillance system for foodborne pathogens.

PathOGiST: A Novel Method for Clustering Pathogen Isolates by Combining Multiple Genotyping Signals

In this paper we study the problem of clustering bacterial isolates into epidemiologically related groups from next-generation sequencing data. Existing methods for this problem mainly use a single genotyping signal, and either use a distance-based method with a pre-specified number of clusters, or a phylogenetic tree-based method with a pre-specified threshold. We propose PathOGiST, an algorithmic framework for clustering bacterial isolates by leveraging multiple genotypic signals and calibrated thresholds. PathOGiST uses different genotypic signals, clusters the isolates based on these individual signals with correlation clustering, and combines the clusterings based on the individual signals through consensus clustering. We implemented and tested PathOGiST on three different bacterial pathogens - Escherichia coli, Yersinia pseudotuberculosis, and Mycobacterium tuberculosis - and we conclude by discussing further avenues to explore.

A Simple and Robust Statistical Method to Define Genetic Relatedness of Samples Related to Outbreaks at the Genomic Scale - Application to Retrospective Salmonella Foodborne Outbreak Investigations

The investigation of foodborne outbreaks (FBOs) from genomic data typically relies on inspecting the relatedness of samples through a phylogenomic tree computed on either SNPs, genes, kmers, or alleles (i.e., cgMLST and wgMLST). The phylogenomic reconstruction is often time-consuming, computation-intensive and depends on hidden assumptions, pipelines implementation and their parameterization. In the context of FBO investigations, robust links between isolates are required in a timely manner to trigger appropriate management actions. Here, we propose a non-parametric statistical method to assert the relatedness of samples (i.e., outbreak cases) or whether to reject them (i.e., non-outbreak cases). With typical computation running within minutes on a desktop computer, we benchmarked the ability of three non-parametric statistical tests (i.e., Wilcoxon rank-sum, Kolmogorov-Smirnov and Kruskal-Wallis) on six different genomic features (i.e., SNPs, SNPs excluding recombination events, genes, kmers, cgMLST alleles, and wgMLST alleles) to discriminate outbreak cases (i.e., positive control: C+) from non-outbreak cases (i.e., negative control: C-). We leveraged four well-characterized and retrospectively investigated FBOs of Salmonella Typhimurium and its monophasic variant S. 1,4,[5],12:i:- from France, setting positive and negative controls in all the assays. We show that the approaches relying on pairwise SNP differences distinguished all four considered outbreaks in contrast to the other tested genomic features (i.e., genes, kmers, cgMLST alleles, and wgMLST alleles). The freely available non-parametric method written in R has been designed to be independent of both the phylogenomic reconstruction and the detection methods of genomic features (i.e., SNPs, genes, kmers, or alleles), making it widely and easily usable to anybody working on genomic data from suspected samples.