# Diagnostic sensitivity and specificity of metagenomic sequencing and qPCR for detection of viruses associated with bovine respiratory disease estimated using Bayesian latent class models

## TL;DR

This study compares the accuracy of two methods for detecting viruses in cattle before they show symptoms of respiratory disease.

## Contribution

The study introduces Bayesian latent class modeling to evaluate diagnostic tests without a gold standard in bovine respiratory disease detection.

## Key findings

- qPCR had higher sensitivity than sequencing for BCoV and BoHV-1, but sequencing was better for BRSV.
- Diagnostic specificity for BCoV was higher with sequencing than with qPCR.
- Both methods showed moderate to low sensitivity for most BRD viruses.

## Abstract

Very few studies have examined the diagnostic sensitivity and specificity of currently available laboratory tests for detecting respiratory pathogens in cattle, and even fewer have examined test performance on samples from animals before the onset of clinical disease.

In this study, Bayesian latent class modeling (BLCM) was used to assess diagnostic test performance in the absence of a gold standard on nasal swabs collected from 19 western Canadian feedlots. Viruses associated with bovine respiratory disease (BRD) were identified using qPCR from a commercial diagnostic laboratory from 760 nasal swabs collected from fall-placed calves (FPC) and yearlings (YRL) at and shortly after feedlot arrival. Using BLCM, the qPCR results were compared to previously reported matching nanopore metagenomic sequencing data for these same samples. Based on BLCM, test sensitivities and specificities were estimated for the detection of bovine coronavirus (BCoV), bovine herpesvirus type 1 (BoHV-1), bovine parainfluenza virus type 3 (BPIV-3), bovine respiratory syncytial virus (BRSV), and influenza D virus (IDV). Estimates informed by BLCM were not available for the detection of bovine viral diarrheal virus (BVDV) because qPCR did not detect this virus in any samples.

Diagnostic sensitivity of qPCR was higher than metagenomic sequencing for detecting BCoV (qPCR 0.90, 95% CrI 0.81–0.99; sequencing 0.35, 95% CrI 0.25–0.46) and BoHV–1 (qPCR 0.39, 95% CrI 0.19–0.99; sequencing 0.04, 95% CrI 0.01–0.15). However, the estimated diagnostic sensitivity of metagenomic sequencing was higher than qPCR for identifying BRSV (qPCR 0.32, 95% CrI 0.22–0.43; sequencing 0.60, 95% CrI 0.44–0.77). No significant difference among sensitivities was noted for the detection of BPIV-3 (qPCR 0.42, 95% CrI 0.21–0.66; sequencing 0.52, 95% CrI 0.19–0.87) and IDV (qPCR 0.65, 95% CrI 0.53–0.79; sequencing 0.60, 95% CrI 0.48–0.73). Diagnostic specificity was comparable for most viruses, except for BCoV, where metagenomic sequencing (BCoV 0.91, 95% CrI 0.88–0.95) outperformed qPCR (BCoV 0.59, 95% CrI 0.51–0.68). The specificity and sensitivity for detection of BRD-associated bacteria from the same metagenomic data were also similar to those estimated for culture and qPCR results for the same samples.

Estimated test sensitivities of both nanopore metagenomic sequencing and qPCR for the detection of BRD viruses of interest in nasal swab samples were moderate to very low for most viruses. While the tests varied in their ability to detect individual viruses, data from this study suggest nanopore metagenomic sequencing offers a potential alternative for diagnostic laboratories to identify three of six important BRD viruses as well as bacteria associated with BRD.