Detecting Disease from Breath: How 'Breathomics' is Turning the Corner

Editor's note: This is part one of a two-article series on breathomics, as published in the February print issue of Laboratory Equipment.

Researchers and doctors have long investigated the connection between a person’s health and their exhaled breath. As far back as 400 B.C., Hippocrates told his students to smell their patients’ breath to search for disease clues, such as diabetes, which often leaves a sweet smell on the breath of its host.

DEEPER INSIGHTS

A method for the Analysis and Quantitation of Cocaine on Currency

Fast forward to 2017 and breath research—or breathomics, as its now called—is just as important as ever. The field is a combination of breath and metabolomics, the study of unique chemical fingerprints that specific cellular processes leave behind. For example, the altered cellular metabolism of tumors and other diseased tissues yields different chemical “fingerprints” of volatile waste products in breath exhaled by a patient compared with a healthy individual.

Once relegated to the lab due to the mass spectrometric techniques that traditionally powered breathomic tests, the field is now emerging for some fresh air, with researchers increasingly developing portable breathomic devices for everything from diabetes to cancer.

Diabetes detection

Mass spectrometry’s ability to detect minute traces of compounds is well suited for the field of breathomics. But it’s large size and expensive nature render it practically useless for point-of-care testing. Researchers from the University of Oxford and Oxford Medical Diagnostics strive to fill the gap between these two extremes with a handheld device—suitable for point-of-care testing—that can identify when a patient has undiagnosed Type I diabetes, or has problems controlling their blood glucose.

According to the researchers, their device collects breath for sampling from a storage bag, or directly via the unit. A pump draws the sample through a tiny preconcentrator that traps the breath acetone but allows other breath constituents, such as methane, water and carbon dioxide, to flow through. This is the step that has often been a challenge for portable breathalyzers since breath contains a complex mix of compounds that can skew results if not using a sensitive mass spectrometer. But, the researchers were able to overcome this limitation.

Their device utilizes an absorbent polymer called Porapak—the same one used in gas chromatography columns—as the material for the preconcentrator. Once the preconcentrator is heated, it releases the acetone into a sensitive optical sensor that detects at a wavelength specific to acetone.

Robert Peverall and colleagues tested the accuracy of their device on healthy volunteers who had undergone different periods of exercise and fasting, which can lead to an increase in breath acetone. The breathalyzer’s results were compared with and verified against data obtained using a soft ionization mass spectrometer. The measurements were a close match and covered a wide range of concentrations, including those that would suggest a patient has undiagnosed Type I diabetes.

Peverall told Laboratory Equipment it wouldn’t take much to turn the breathalyzer into a functional point-of-care device at this point—all that’s needed is a practical way of changing the consumable materials, which is something basic industrial design could easily accomplish.

“I predict that within the next 10 to 20 years, measuring one or a few biomarkers in breath will be ubiquitous within primary care, and giving breath samples will be just as common as having our blood pressure measured today,” Peverall said.

Seventeen other diseases

Hossam Haick and his collaborators from 14 clinical departments worldwide took a slightly different—or rather, backward—approach to developing a disease breathalyzer. Instead of designing the device first, Haick’s team instead focused on a disease’s unique breathprint by developing an array of nanoscale sensors that could detect the individual volatile organic components associated with specific illnesses. Their analysis identified 13 exhaled chemicals that could differentiate between 17 different diseases. By analyzing the results with artificial intelligence techniques, the team could use the array to diagnose and classify each disease, even detecting multiple diseases from the same breath sample.

“The system is inspired by and mimics the mammalian sense of smell,” Haick explained to Laboratory Equipment. “In order to achieve these results, we used advanced pattern recognition algorithms, working similarly to the olfactory cortex, where electrical signals from nerves are translated into smells and eventually recognized by learning procedure in the memory cortex. Therefore, we used the signals of the sensors exposed to ‘training sets’ of breath samples to recognize the chemical patterns of each, train the system and create artificial memory of the signatures.”

The performance of the artificially intelligent nanoarray was clinically assessed in blind experiments that showed an 86 percent accuracy between the nanoarray and the lab-standard gas chromatography mass spectrometry (GCMS).

Other researchers have tried, and failed, to support the hypothesis that a single VOC can discriminate between different diseases. What makes Haick’s research different is the application of A.I. and mathematical models—these support the study’s finding that the use of VOC patterns in exhaled breath is a realistic option for discriminating between different disease states.

While larger translational studies are required to further validate Haick’s findings, the research definitely points toward a method for developing inexpensive, easy-to-use, miniaturized tools for personalized screening and diagnosis of a range of diseases.

“For full and efficient operation in real clinical settings, the artificially intelligent nanoarray needs to be increasingly trained using known clinical samples to build up a consistent and reliable database of reference,” reads Haick’s paper, published recently in ACS Nano. “It can then recognize new samples by comparing disease-related VOC patterns to those already in its database.”

In addition, next-generation sensors could be developed and modified to suit particular VOC compositions, providing a sensor that is highly sensitive to individual VOC profiles.