Comparison of the performance of two targeted metagenomic virus capture probe-based methods using reference control materials and clinical samples

Abstract

Viral enrichment by probe hybridization has been reported to significantly increase the sensitivity of viral metagenomics. This study compares the analytical performance of two targeted metagenomic virus capture probe-based methods: (i) SeqCap EZ HyperCap by Roche (ViroCap) and (ii) Twist Comprehensive Viral Research Panel workflow, for diagnostic use. Sensitivity, specificity, and limit of detection were analyzed using 25 synthetic viral sequences spiked in increasing proportions of human background DNA, eight clinical samples, and American Type Culture Collection (ATCC) Virome Virus Mix. Sensitivity and specificity were 95% and higher for both methods using the synthetic and reference controls as gold standard. Combining thresholds for viral sequence read counts and genome coverage [respectively 500 reads per million (RPM) and 10% coverage] resulted in optimal prediction of true positive results. Limits of detection were approximately 50-500 copies/mL for both methods as determined by ddPCR. Increasing proportions of spike-in cell-free human background sequences up to 99.999% (50 ng/mL) did not negatively affect viral detection, suggesting effective capture of viral sequences. These data show analytical performances in ranges applicable to clinical samples, for both probe hybridization metagenomic approaches. This study supports further steps toward more widespread use of viral metagenomics for pathogen detection, in clinical and surveillance settings using low biomass samples.

Importance: Viral metagenomics has been gradually applied for broad-spectrum pathogen detection of infectious diseases, surveillance of emerging diseases, and pathogen discovery. Viral enrichment by probe hybridization methods has been reported to significantly increase the sensitivity of viral metagenomics. During the past years, a specific hybridization panel distributed by Roche has been adopted in a broad range of different clinical and zoonotic settings. Recently, Twist Bioscience has released a new hybridization panel targeting human and animal viruses. This is the first report comparing the performance of viral metagenomic hybridization panels.

Keywords: capture probes; targeted metagenomics; viral diagnostics; viral metagenomics.

FIG 1

Workflows of the capture probe-based targeted metagenomic protocols compared in this study, Twist Comprehensive Viral Research, and the SeqCap EZ HyperCap (ViroCap, Roche) and, both in combination with identical bioinformatic analyses pipeline. Created using BioRender.

FIG 2

ROC curves for the prediction of detection of viral sequences using the virus capture probe-based metagenomic workflows Twist Comprehensive Viral Research, and SeqCap EZ HyperCap (Roche). The validation panel consisted of synthetic viral sequences spiked in a background of human cell-free DNA (90%–99.92%) and diluted ATCC Virome virus mix standard (copies/mL ranging from 10⁴ to 10⁷). (A) ROC based on varying threshold of sequence RPM, and (B), based on varying threshold of percentage of genome coverage for defining a positive result, using a random selection of 1 million sequence reads per data set. For all curves (A and B), a minimum of three distributed regions of the genome covered was set as primary parameter for defining detection.

FIG 3

Correlation graph depicting linearity between the VL (log₁₀ IU and C/mL, horizontally) and the log₁₀ RKPM genome as generated using the virus capture probe-based metagenomic workflows Twist Comprehensive Viral Research and SeqCap EZ HyperCap (Roche). Included are detections by both methods from synthetic viral sequences spiked in a background of human cell-free DNA (90–99.991%), dilution series (see Fig. 4), and clinical samples.

FIG 4

LOD of viral sequences using the virus capture probe-based metagenomic workflows Twist Comprehensive Viral Research (depicted in the left upper corner, “C”), and SeqCap EZ HyperCap (Roche, depicted in the right lower corner, “H”). RPKM genome are shown for different VLs (c or IU/mL, see Table S1). The samples consisted of synthetic viral sequences spiked in a background of human cell-free DNA (90–99.99990%) (Inf A, SARS-CoV-2, HBoV), and clinical EDTA plasma samples (ADV, EBV, HBV). NT, not tested. Created using BioRender.

FIG 5

Reproducibility of read counts and genome coverage percentages. Between-run variability in three samples using the parameters RPKM (bars, left axis) and genome coverage percentage (black dots, right axis) as generated using the virus capture probe-based metagenomic workflows Twist Comprehensive Viral Research and SeqCap EZ HyperCap (Roche). The percentage of genome coverage was based on a random selection of 1 million sequence reads per data set. Coefficients of variance in RPKM ranged from 0.0% to 4.7% (see Table S1). Created using BioRender.