Towards standardisation: comparison of five whole genome sequencing (WGS) analysis pipelines for detection of epidemiologically linked tuberculosis cases.
Jajou R., Kohl TA., Walker T., Norman A., Cirillo DM., Tagliani E., Niemann S., de Neeling A., Lillebaek T., Anthony RM., van Soolingen D.
BackgroundWhole genome sequencing (WGS) is a reliable tool for studying tuberculosis (TB) transmission. WGS data are usually processed by custom-built analysis pipelines with little standardisation between them.AimTo compare the impact of variability of several WGS analysis pipelines used internationally to detect epidemiologically linked TB cases.MethodsFrom the Netherlands, 535 <i>Mycobacterium tuberculosis</i> complex (MTBC) strains from 2016 were included. Epidemiological information obtained from municipal health services was available for all mycobacterial interspersed repeat unit-variable number of tandem repeat (MIRU-VNTR) clustered cases. WGS data was analysed using five different pipelines: one core genome multilocus sequence typing (cgMLST) approach and four single nucleotide polymorphism (SNP)-based pipelines developed in Oxford, United Kingdom; Borstel, Germany; Bilthoven, the Netherlands and Copenhagen, Denmark. WGS clusters were defined using a maximum pairwise distance of 12 SNPs/alleles.ResultsThe cgMLST approach and Oxford pipeline clustered all epidemiologically linked cases, however, in the other three SNP-based pipelines one epidemiological link was missed due to insufficient coverage. In general, the genetic distances varied between pipelines, reflecting different clustering rates: the cgMLST approach clustered 92 cases, followed by 84, 83, 83 and 82 cases in the SNP-based pipelines from Copenhagen, Oxford, Borstel and Bilthoven respectively.ConclusionConcordance in ruling out epidemiological links was high between pipelines, which is an important step in the international validation of WGS data analysis. To increase accuracy in identifying TB transmission clusters, standardisation of crucial WGS criteria and creation of a reference database of representative MTBC sequences would be advisable.