Genomic variant-identification methods may alter Mycobacterium tuberculosis transmission inferences

Abstract

Pathogen genomic data are increasingly used to characterize global and local transmission patterns of important human pathogens and to inform public health interventions. Yet, there is no current consensus on how to measure genomic variation. To test the effect of the variant-identification approach on transmission inferences for Mycobacterium tuberculosis, we conducted an experiment in which five genomic epidemiology groups applied variant-identification pipelines to the same outbreak sequence data. We compared the variants identified by each group in addition to transmission and phylogenetic inferences made with each variant set. To measure the performance of commonly used variant-identification tools, we simulated an outbreak. We compared the performance of three mapping algorithms, five variant callers and two variant filters in recovering true outbreak variants. Finally, we investigated the effect of applying increasingly stringent filters on transmission inferences and phylogenies. We found that variant-calling approaches used by different groups do not recover consistent sets of variants, which can lead to conflicting transmission inferences. Further, performance in recovering true variation varied widely across approaches. While no single variant-identification approach outperforms others in both recovering true genome-wide and outbreak-level variation, variant-identification algorithms calibrated upon real sequence data or that incorporate local reassembly outperform others in recovering true pairwise differences between isolates. The choice of variant filters contributed to extensive differences across pipelines, and applying increasingly stringent filters rapidly eroded the accuracy of transmission inferences and quality of phylogenies reconstructed from outbreak variation. Commonly used approaches to identify M. tuberculosis genomic variation have variable performance, particularly when predicting potential transmission links from pairwise genetic distances. Phylogenetic reconstruction may be improved by less stringent variant filtering. Approaches that improve variant identification in repetitive, hypervariable regions, such as long-read assemblies, may improve transmission inference.

One is not enough: On the effects of reference genome for the mapping and subsequent analyses of short-reads

“…They also reflect the correlation between genetic distance of isolates to the reference genome and the number of called variants which, as highlighted in previous studies, could be associated with the increase of false positives when the precision of SNP calling decreases [ 23 , 28 , 42 ]. However, the increase in false positive rates in variant detection may be affected not only by distance between query and reference genomes (measured as SNP distance), but also by genomic architecture or assembly quality [ 56 ]. This suggests that multiple aspects should be considered when choosing a reference for mapping, aside from SNP distance between reads and reference genome.…”

The role of prisons in disseminating tuberculosis in Brazil: a genomic epidemiology study

“…(31,32) Future studies that leverage within-host M. tuberculosis diversity and/or include the diverse PE/PPE genes could provide greater resolution and reduce the uncertainty of transmission inferences. (33) In this study, we present genomic evidence that prisons act as tuberculosis reservoirs in Central West, Brazil. The dramatic expansion of incarceration in recent decades has put an increasing population at extremely high risk of tuberculosis; this risk extends to surrounding communities.…”

A Bioinformatics Whole-Genome Sequencing Workflow for Clinical Mycobacterium tuberculosis Complex Isolate Analysis, Validated Using a Reference Collection Extensively Characterized with Conventional Methods and In Silico Approaches

“…AMR) (67,68), or detecting the correct sequence type for MLST (44). Walter et al recently used a set of 85 SNPs confirmed with Sanger sequencing to evaluate the performance of various variant calling tools for MTBC outbreak investigation (71). Even though individual variants were called with relatively high accuracy, inconsistencies between tools impacted transmission inferences.…”