P034
Comparison of Illumina and Oxford Nanopore Technology sequencing of Mycobacterium tuberculosis
M Colpus(1) C Baker(1) H N Hong(3) S V Omar(2) P P Trieu(3) D DA Thu(3) L Hall(1) P W Fowler(1) T Walker(3,4) R Spies(1) H Webster(1) J Westhead(1) H Thai(1) R Turner(1) T Peto(1) N L Quang(3) N TT Thuong(3,4) D Crook(1)
1:Nuffield Department of Medicine, University of Oxford; 2:National Institute for Communicable Diseases, South Africa; 3:Oxford University Clinical Research Unit, Centre for Tropical Medicine, Ho Chi Minh city, Vietnam; 4:Nuffield Department of Medicine, Centre for Tropical Medicine and Global Health, University of Oxford, Oxford, England, UK
Illumina sequencing is the leading sequencing platform for Mycobacterium tuberculosis. It has become the standard for determining genomic variation, classifying subspecies, predicting drug resistance, and identifying outbreak clusters.
There have been only a handful of studies comparing the relative performance of long-read sequencing by Oxford Nanopore Technologies (ONT) to Illumina. Here we investigate the differences in performance of 400 genomes sequenced by both Illumina and ONT and initially report on the first 67. Half were South African and half Vietnamese. Sequencing followed standard operating procedures. Genomes were assembled and analysed using an online cloud platform, which found general concordance but also clear differences between ONT and Illumina. Ongoing characterisation and quantification of differences will be reported for all 400.
Illumina samples had a mean total bases 0.67Gbp. ONT produced 0.51Gbp with a mean N50 of 3670bp. 64/67 genomes were assigned identical lineages; 3 samples had a mixed lineage in only one platform. Resistance prediction for 61/67 genomes was identical (across 14 antibiotics), with discrepancies linked to indels with insufficient read support.
For each isolate we calculated the SNP distance between sequencing platforms. Repetitive regions were excluded using masks. The mean difference was 5.6 SNPs. We examined whether platform specific errors or difference in mapping caused this discrepancy and also investigated mask refinement for improving cross-platform detection of clusters. To further aid investigation of platform differences, ONT de novo assembly was performed yielding assemblies of expected total length (12 circularised and 48 consisting of 1-4 contigs).