Title: Diagnosis and Classification of 17 Diseases from 1404 Subjects via Pattern Analysis of Exhaled Molecules

Authors: Hossam Haick et al.

Publication Info: ACS Nano, 2017, 11 (1), pp 112–125 DOI: 10.1021/acsnano.6b04930

This work reports on the development of a portable, inexpensive, and non-invasive breath analyzer that couples the use basic fundamental chemical principles with pattern recognition algorithms to simultaneously detect and discriminate between 17 different diseases  including various forms of cancer and Parkinson’s disease.

Since antiquity breath odor has been employed to assess human health. Through normal metabolic activity endogenous volatile organic compounds are released into the bloodstream and subsequently excreted through exhalation.  As breath analysis has evolved over the years, specific relative combinations of certain exhaled organic compounds have been identified as biomarkers for various health conditions.  

While breath analysis has become a valid health diagnostic tool, its ubiquity is currently hindered by a few aspects that this work seeks to overcome.  It typically requires patients to visit clinics  staffed with trained technicians to perform the diagnostic and maintain, calibrate, and operate costly and sensitive instrumentation – typically a mass spectrometer coupled to a gas chromatography implement.  The inexpensive and disposable breath sensors proposed in this work will enable health assessment in the convenience of one’s own home – like a home pregnancy test or blood insulin level test.   

The breath analyzers for this work were fabricated on inexpensive silicon wafers and the entire footprint of a sensor is about the size of a fingernail.  The actual sensing component is an array of 3 millimeter diameter discs made conductive through drop-casting a small quantity of gold nanoparticles or carbon nanotubes from solution.  Each disc in the array is made unique through molecular modification with different small molecules such as  dodecanethiol, chlorobenzene, or methanethiol.

Various electrical responses (such as resistance, impedance, and time-resolved resistance and impedance) measured across each modified pad provide the sensing outputs of the breath analyzer.  When a sample of breath passes over the conductive pads, the various organic compounds present can induce measurable changes in the electrical properties of the conductive pads.  In response to the types, extent, and timescales of basic intermolecular forces (i.e. dispersion, dipole-dipole, and hydrogen bonding interactions) that occur between  the endogenous organic compounds exhaled in the breath and the molecular modifiers on the conductive pads, the electrical characteristics of the pads can be altered and readily measured.  For example, significant sorption of a ketone in the breath may occur at a pad functionalized with an amide group.  Inclusion of this ketone analyte may cause swelling of the molecular layers which function to bridge the individual gold nanoparticles, and thus changing the electrical properties (e.g. the conductivity, dielectric constant, impedance, etc.) of that specific pad over a period of time until the analyte desorbs back into the flowing air stream.

The analysis component of this sensor is an interesting attempt at biomimicry of the human olfactory system.  The human olfactory system is composed of ~350 different olfactory receptors and and each of these receptor types exhibit subtle differences in composition and structure. None of these receptors are exclusively selective for any single odorant (e.g. there is not a specific receptor for cinnamaldehyde). Each receptor type actually  exhibits relatively poor selectivity and have evolved compositional and structural features that enable a wide range of intermolecular interactions to occur between it and the odorant.  An electrical response based on the strength and dynamics of the receptor-odorant interactions is conducted to the brain from the array of olfactory neurons. It is the simultaneous combination of electrical responses of many neurons, not a single neuron, that allow the brain to amazingly identify a specific odorant.  Smell does not occur in the nose, but in the brain!

In this sensor platform, the array of olfactory neurons composing the human olfactory system are synthesized by the uniquely functionalized discs, and the brain functionality  is synthesized by a software artificial neural network that is trained with breath samples of patients with previously known health status.  While this report emphasis the pattern recognition algorithms and information processing employed to detect disease biomarkers in complex breath samples, the root of this work is in the fundamental basic chemistry that enables it all – intermolecular interactions between the functionalized pads and various organic compounds in the breath.