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

Abstract

The use of whole genome sequencing (WGS) for routine typing of bacterial isolates has increased substantially in recent years. For Mycobacterium tuberculosis (MTB), in particular, WGS has the benefit of drastically reducing the time to generate results compared to most conventional phenotypic methods. Consequently, a multitude of solutions for analyzing WGS MTB data have been developed, but their successful integration in clinical and national reference laboratories is hindered by the requirement for their validation, for which a consensus framework is still largely absent. We developed a bioinformatics workflow for (Illumina) WGS-based routine typing of MTB Complex (MTBC) member isolates allowing complete characterization including (sub)species confirmation and identification (16S, csb/RD, hsp65), Single Nucleotide Polymorphism (SNP)-based antimicrobial resistance (AMR) prediction, and pathogen typing (spoligotyping, SNP barcoding, and core genome MultiLocus Sequence Typing). Workflow performance was validated on a per-assay basis using a collection of 238 in-house sequenced MTBC isolates, extensively characterized with conventional molecular biology-based approaches supplemented with public data. For SNP-based AMR prediction, results from molecular genotyping methods were supplemented with in silico modified datasets allowing to greatly increase the set of evaluated mutations. The workflow demonstrated very high performance with performance metrics >99% for all assays, except for spoligotyping where sensitivity dropped to ∼90%. The validation framework for our WGS-based bioinformatics workflow can aid standardization of bioinformatics tools by the MTB community and other SNP-based applications regardless of the targeted pathogen(s). The bioinformatics workflow is available for academic and non-profit usage through the Galaxy instance of our institute at https://galaxy.sciensano.be.

Evaluation of WGS performance for bacterial pathogen characterization with the Illumina technology optimized for time-critical situations

“…This effectively entails that high performance at reduced sequencing depths and/or coverages can be potentially misleading if performance of the bioinformatics assay for the full dataset is poor to begin with. Notwithstanding, this effect is expected to be minimal in our study because we specifically used bioinformatics workflows already rigorously validated previously to be of very high performance [31][32][33]. While we employed 95 % as a threshold to consider results as high-quality, higher cut-offs can be enforced when quality needs to be as close to that of the full dataset as possible.…”

“…Confirmation that the targeted species is present trough alignment of the assembled contigs against the NCBI 16S rRNA database [ 33 ]…”

“…We built upon our previously described validation strategy to evaluate the effect of decreasing read lengths, coverages, and their interaction, on the performance of various bioinformatics assays for three different species [31][32][33]. An overview and short descriptions of the evaluated assays are provided in Table 1, along with references to their corresponding publications describing in detail the employed bioinformatics methodology.…”

“…Here, we evaluated both decreasing overall run time by reducing read lengths, and analysing isolate WGS data in real-time from Illumina instruments while the sequencer is still running, enabling partially generated datasets to be fed into bioinformatics workflows for typing and characterization of pathogenic isolates when response time is essential. We performed an exhaustive performance evaluation at decreasing read lengths and coverages in function of sequencing run time to provide specific recommendations for guaranteeing a minimum required level of performance based on a previously described validation strategy [31–33], using high-quality reference datasets for three pathogens ( Neisseria meningitidis , Mycobacterium tuberculosis , and Shiga-toxin producing Escherichia coli ), and evaluating multiple bioinformatics assays of public health interest that can also be adapted for other time-critical WGS applications. Specific recommendations to set up time-optimized experiments are provided through the characterization of assay performance in function of read lengths, coverage, and estimated sequencing duration.…”

Consistency of Mycobacterium tuberculosis Complex Spoligotyping between the Membrane-Based Method and In Silico Approach

“…It is well known that the initial amplification step of the former (that is not required for the latter) may introduce bias in the output sequences mainly due to heterogeneous or variant-sensitive primer affinity ( 13 ). In a recent study, Bogaerts et al ( 14 ) compared these methods using 166 MTBC from the Belgian National Reference Center, but the discrepancies were not explored. In the present study, we compared WGS-based MTBC identification and spoligotyping (43 spacers) with those obtained by membrane hybridization assays in a French cohort.…”