Compatibility of whole-genome sequencing data from Illumina and Ion Torrent technologies in genome comparison analysis of Listeria monocytogenes

TL;DR

This study compares genome data from two sequencing technologies for Listeria monocytogenes and finds that while compatibility can be improved, it is not perfect.

Contribution

The study evaluates compatibility of Illumina and Ion Torrent sequencing data in genome comparison of Listeria monocytogenes, revealing discrepancies in cgMLST.

Findings

Only SPAdes assembler delivered comparable results between sequencing platforms.

cgMLST allele discrepancy averaged 14.5 alleles, exceeding the 7-allele threshold for clustering.

Strict frameshift filtering reduced discrepancy but lowered discriminatory power.

Abstract

Whole-genome sequencing (WGS) has become the key approach for molecular surveillance of Listeria monocytogenes. Genome comparison analysis can reveal transmission routes that cannot be found with classic epidemiology. A widespread standard for use in genome comparison analysis involves data from short-read sequencing, generated on Illumina or Ion Torrent devices. To date, little is known about the compatibility of data from both platforms. This knowledge is essential when it comes to the central analysis of data, for example, in the case of outbreaks. We used WGS data from 47 L. monocytogenes isolates of the strain collection of the German National Reference Laboratory for L. monocytogenes, generated on either Illumina or Ion Torrent devices, to analyse the impact of the sequencing technology on downstream analyses. In our study, only the assembler SPAdes delivered qualitatively comparable results. In the gene-based core genome multilocus sequence typing (cgMLST), the same-strain allele discrepancy between the platforms was 14.5 alleles on average, which is well above the threshold of 7 alleles routinely used for cluster detection in L. monocytogenes. An application of a strict frameshift filter in cgMLST analysis could push the mean discrepancy below this threshold but reduced discriminatory power. The impact of the platform on the read-based single nucleotide polymorphism analysis was lower than that on the cgMLST. Overall, it was possible to improve compatibility in various ways, but perfect compatibility could not be achieved.

Figures