COVID-19 detection through scent analysis with a compact GC device

Recent studies, including ours, have suggested that breath may allow us to diagnose COVID-19 infectionand even monitor its progress. As compared to immunological and genetic based methods using sample medialike blood, nasopharyngeal swab, and saliva, breath analysis is non-invasive, simple, safe, and inexpensive; itallows a nearly infinite amount of sample volume and can be used at the point-of-care for rapid detection.Fundamentally, breath also provides critical metabolomics information regarding how human body responds tovirus infection and medical intervention (such as drug treatment and mechanical ventilation). The objectives ofthe proposed SCENT project are: (1) to refine automated, portable, high-performance micro-gaschromatography (GC) device and related data analysis / biomarker identification algorithms for rapid (5-6minutes), in-situ, and sensitive (down to ppt) breath analysis and (2) to conduct breath analysis on up to 760patients, and identify and validate the COVID-19 biomarkers in breath. Thus, in coordination with the RADx-radData Coordination Center (DCC), we will complete the following specific aims.(1) Refine 5 automated micro-GC devices to achieve higher speed and better separation capability. Wewill construct 5 new automated and portable one-dimensional micro-GC devices that require only ~6 minutes ofassay time (improved from current 20 minutes) at the ppt level sensitivity (Sub-Aim 1a). Then the devices will beupgraded to 2-dimensional micro-GC to significantly increase the separation capability (Sub-Aim 1b). In themeantime, we will optimize and automate our existing data processing and biomarker identification algorithmsand codes to streamline the workflow so that the GC device can automatically process and analyze the datawithout human intervention (Sub-Aim 1c).(2) Identify breath biomarkers that distinguish COVID-19 positive (symptomatic and asymptomatic) andnegative patients. We will recruit a training cohort of 380 participants, including 190 COVID-19 positive patients(95 symptomatic and 95 asymptomatic) and 190 COVID-19 negative patients from two hospitals (MichiganMedicine - Ann Arbor and the Henry Ford Hospital - Detroit). We will conduct breath analysis using machinelearning to identify VOC patterns that match each COVID-19 diagnostic status.(3) Validate the COVID-19 biomarkers using our refined micro-GC devices. Using the refined 2-D micro-GCdevices from Sub-Aim 1b, we will recruit a new validation cohort of 380 participants (190 COVID-19 positivepatients and 190 COVID-19 negative patients) to validate the biomarkers identified in Aim 2. We will leverage existing engineering, data science, clinical, regulatory, and commercialization resourcesthroughout the project to hit our milestones, ensuring a high likelihood of rapid patient impact. Upon completionof this work, we will have a portable micro-GC device and accompanying automated algorithms that can detectand monitor COVID-19 status for people in a variety of clinical and community settings.