Odorprint Based Disease Diagnostics - Project Summary
It has long been observed that certain diseases may be diagnosed by smell alone. There is mounting evidence
supporting these observations, showing that the metabolic changes brought about by disease, expressed in
biospecimens such as sweat, breath, urine and blood, can be accurately identified through olfaction. This is the
case not only for metabolic diseases such as diabetes, but most notably cancer, Alzheimer’s, Parkinson’s, and
many types of infection, including COVID-19. But it remains a mystery how olfactory systems achieve this ability,
especially when faced with the stark levels of variance in healthy populations, and the challenge of identifying a
complex odor object against irrelevant background components. This project will investigate the neural
mechanisms of odor-based disease diagnostics in the olfactory system of the mouse.
Initial experiments will image the responses of olfactory sensory neurons in the olfactory bulb of the awake
mouse. Using mouse models of disease, we will collect urine samples corresponding to both disease and healthy
states, with controlled between-sample variability. We will image glomeruli, with each glomerulus aggregating
the axons of sensory neurons expressing the same class of receptor. Linear and nonlinear dimensionality
reduction methods will be developed to analyze the complex spatiotemporal patterns of glomerular activity
elicited by disease and healthy control samples. From this analysis, the key features of neural activity that
underpin disease detection will be identified, and related to specific glomeruli. Glomeruli of interest will then be
used to isolate the volatile organic compounds of relevance, through gas chromatography-olfactometry in parallel
with gas-chromatography/mass-spectrometry. Additionally, quantitative methods will be developed for the
alignment of neural spaces across multiple mice, using a minimal number of odors. This will render odor features
translatable across animals, allowing for the decoding of disease in mice without extensive training data
collection. The developed experimental and computational pipeline will be then applied to detect and decipher
odorprints of multiple human diseases.
Understanding how olfactory systems detect disease has the potential to revolutionize medical diagnostics,
particularly with respect to early and noninvasive screening. But it will also constitute progress in ‘cracking the
olfactory code’, with our understanding of olfaction currently lagging behind vision and audition. From an
evolutionary perspective, the natural stimuli of olfaction were the metabolic states of food, mates, peers, and
predators, rarely the monomolecular odorants commonly used in olfaction research today. While this project has
an applied aim of medical diagnostics, the path to that aim proceeds via a deep understanding of some of the
fundamental, yet still mysterious, principles of olfaction.