# Trained Scent Dog Detection and GC-MS Analysis of Volatile Organic Compounds from Murine Coronavirus-Infected Cell Cultures

## TL;DR

This study shows that trained dogs can detect coronavirus-infected cells by scent, supported by chemical analysis of volatile compounds.

## Contribution

The study introduces a validated framework for using scent dogs and VOC analysis for non-invasive viral detection.

## Key findings

- Trained dogs identified infected samples with high sensitivity (0.95) compared to controls.
- 3-heptanone and 1-nonanol were unique to infected samples, while others were significantly elevated.
- A machine learning model achieved 0.82 accuracy using VOC profiles from infected samples.

## Abstract

Detecting viral infections quickly and accurately is critical for protecting public and animal health. In this study, we explored whether specially trained dogs can detect signs of viral infection by scent alone. We used a harmless murine virus that infects cells in the lab. The dogs were trained to tell the difference between virus-infected and uninfected samples. At the same time, we used a chemical method to identify which specific scent molecules were present. Some of these molecules were only found in infected samples, while others were present at much higher concentrations than in uninfected samples. The dogs were able to correctly identify virus-infected samples in most cases. A computational method using the chemical data also correctly identified infected samples with high accuracy. This work supports the idea that trained dogs can help detect infections through scent and may improve future non-invasive methods for viral detection.

Volatile organic compounds (VOCs) are increasingly recognized as metabolic byproducts of viral infection and may serve as olfactory cues detectable by trained scent dogs. This study examined whether dogs could distinguish cell culture samples infected with murine hepatitis virus strain 1 (MHV-1), a biosafety level 2 coronavirus model, from uninfected controls. Parallel chemical analysis using gas chromatography–mass spectrometry (GC-MS) identified 14 VOCs in infected and 12 in control samples. Notably, 3-heptanone and 1-nonanol were unique to infected samples, while others such as acetophenone, nonanal, decanal, and benzaldehyde were significantly elevated—often by 1.5 to 3 times—in infected cultures. Two trained dogs demonstrated high detection sensitivity (0.95) for infected samples compared to a previously trained odor cinnamon group (0.88) and responded with shorter latency (p = 0.04), suggesting perceptual salience of infection-related VOCs. Reliable detection required pooled volumes (~600 µL), suggesting a threshold effect related to VOC concentration. Additionally, a Random Forest-based machine learning classifier trained on the GC-MS-obtained VOC profiles achieved a cross-validated accuracy of 0.82 (SD = 0.25). These findings suggest that dogs use quantitative VOC differences, rather than unique compounds, for detection. The study provides a validated experimental framework for olfactory diagnostics of viral infections and highlights the potential of scent dogs as non-invasive biosensors in both veterinary and public health contexts.