Background Clinical metagenomics (CMg) offers high-throughput respiratory pathogen detection with a wider range than targeted, probe-dependent diagnostics. Sequencing cost and the challenges of high host biomass in non-invasive samples are barriers to the use of CMg in high-throughput respiratory pathogen detection.
Methods We optimised a long-read sequencing workflow to detect RNA viruses in nasopharyngeal swabs, employing pathogen enrichment and ONT sequencing. As a pre-requisite for agnostic pathogen detection, we first derived quality control (QC) criteria and diagnostic thresholds against a gold-standard comprising 23 pathogen targets detected by routine multiplex PCR. We validated this workflow using 344 prospectively collected upper respiratory tract samples submitted for routine testing.
Findings Using pre-defined QC and positivity criteria, the workflow's sensitivity versus PCR was 51% (95%CI: 45%-57%) (133/260 positive targets detected) (ranging from 19%-85% across pathogens with >20 gold- standard detections), and specificity 99.8% (95%CI: 99.6%-99.9%) (3836/3845 negative targets not detected). Sensitivity improved to 58% using post-hoc optimised thresholds, 61% only considering RNA pathogens, 70% excluding rhinovirus/enterovirus and 83% excluding samples with qPCR Ct values ≥35. Read crossover from multiplex sequencing contributed most (7/9) false-positives: only 2 plausible additional pathogens were identified (rhinovirus and coronavirus OC43). 41 respiratory syncytial virus (RSV), 13 influenza A and 10 rhinovirus/enterovirus were successfully sub-typed by sequencing. Multiplexed nanopore sequencing costs were 112GBP/sample.
Interpretation Although CMg has substantial diagnostic potential, this validation study demonstrates the technical limitations of current metagenomic sequencing methods applied to viral detection in upper respiratory tract samples with high host and low pathogen biomass. Its greater sensitivity at higher viral loads demonstrates the importance of identifying the most appropriate use cases to maximise its utility and value.