NEW YORK – University of Minnesota spinout Vocxi Health is preparing to commercialize a point-of-care breath test for lung cancer that relies on highly sensitive biosensors and machine learning.
According to Vocxi CEO Ping Yeh, the company's technology has been in the works for almost a decade, beginning with an innovation challenge at medical device manufacturer Boston Scientific. Lung cancer has proven challenging to detect because a significant portion of biopsies show nodules that aren't cancerous, while some nodules are too small to biopsy. This sparked the question whether a home test for lung cancer could be developed to make lung cancer testing more mobile and accessible, Yeh said, adding he was approached by the technology’s creators to head the new company due to his experience creating other startups.
Working with researchers at the University of Minnesota, the Boston Scientific team aimed to determine if a new class of biosensors would be sensitive enough to pick out volatile organic compounds in the breath.
Breath-based tests have been in development for a while now, often leveraging mass spectrometry. However, the instruments are often large and unwieldy, making them impractical for home or point-of-care use, Yeh said.
In response, Yeh and his colleagues have developed nanosensors and chemistry in a "super-consistent, scalable way to selectively pick out volatile organic compound families to identify patterns" in the breath, he said.
More specifically, they have developed a wafer-based process that creates thousands of chips on a single wafer and a process to lay down a single layer of graphene on the wafers. The researchers also developed a patented process to functionalize and apply a single monolayer of receptor chemistry on top of the graphene, which Yeh said is "critical to the consistency of the sensors."
Simultaneously, machine learning capabilities have grown in the past few years, allowing the company to take the chemical patterns from VOCs, turn them into electrical patterns, and analyze them using machine learning algorithms, Yeh said.
What makes the technology for Vocxi's breath test unique, Yeh noted, is that it doesn't measure just one or a few biomarkers. Instead, it is a "cross-reactive array," meaning that it takes thousands of VOCs in a breath sample, which then cross-react with an array of biosensors, both the graphene-based nanosensors and the monolayer of chemical receptors, he said. The layer of chemical receptors is intended to improve the selectivity of the test.
The chemical patterns detected by the sensors are then converted into electrical patterns, which are fed into a machine learning-based algorithm that analyzes the signals and determines whether lung cancer is present.
According to a study published in 2022 in ACS Nano, each array contains 108 sensors functionalized with 36 chemical receptors for cross-selectivity. Using its supervised machine learning approach, the device showed 98 percent accuracy in detecting five analytes at four concentrations each. The addition of another analyte highly similar to one of the original five dropped the accuracy to 89 percent, the researchers noted.
The company, which was founded in 2022, aims to do more clinical testing as it prepares to commercialize the test, Yeh noted, adding it has done two clinical studies on lung cancer and one on COVID-19, which included 180 participants, that haven't been published yet. He said the clinical studies showed "pretty eye-opening sensitivity and specificity numbers," although he declined to share the exact values. An upcoming multisite pilot study, which Yeh said is likely to begin this summer, is expected to further improve the device's sensitivity and specificity to somewhere in the 90 percent range.
Vocxi plans to begin commercializing its test in the point-of-care setting at doctors' offices and clinics before moving to the home setting, Yeh said. The test will return a risk score that is stratified into a binary outcome to provide a final determination of "high risk," meaning a positive indication of lung cancer, or "low risk," meaning a negative indication for lung cancer.
For home use, a patient would be able to take the test at home and have their results sent immediately to the physician via the cloud for review, Yeh said. This could be particularly useful for patients who are under surveillance for lung nodules that must return to the clinic every few months for lung cancer scans, he added. The current device is connected to a phone, which also makes it well suited for eventual home use.
The firm is also solidifying clinical partnerships with multiple undisclosed partners, he said.
It is also determining its manufacturing and cost structure. The device uses "essentially printed circuit board technology" that is cheap and easy to produce, and its test is reagent-free, so it will be "extremely cost-competitive" against liquid biopsy blood tests or other cancer tests, such as Exact Sciences' Cologuard, which costs about $500, according to Yeh.
Vocxi has just started production of its device, which is about the size of a credit card, and the firm currently has end-to-end production capability. It began manufacturing its commercial hardware design this month and is manufacturing the device both in-house and with long-term partners, Yeh noted. The company is making its devices as needed to support verification, validation, its clinical pilot study, and strategic partnerships, and Vocxi plans to ramp up volumes after the device has been validated and for its larger clinical trial, which will occur after the pilot study.
The intellectual property developed by Boston Scientific and the University of Minnesota has been assigned to the university and is exclusively licensed to Vocxi Health. Last year, the company closed an oversubscribed financing round for an undisclosed amount and is actively fundraising to help with commercialization and additional clinical trials, Yeh said.
Vocxi also has its sights set on other diseases beyond lung cancer, but the hardware of its breath testing device will not change, he noted. Although he declined to specify which other cancers the company will target, Yeh said that the company has just been awarded a grant for researching its test's utility with another cancer type.
According to Raed Dweik, the chief of Cleveland Clinic's Integrated Hospital Care Institute and an expert on breath testing, using breath for disease detection is "not a new thing" and that both mass spectrometry and sensor arrays similar to what Vocxi has developed have pros and cons. Sensor arrays, or "electronic noses," do exactly what they are trained to do, meaning that if the device is trained on a disease it will recognize that disease immediately, but it will not detect something it hasn't been trained on.
In contrast, mass spectrometry will detect every compound in the breath but can't recognize patterns in the way a sensor array can. "You cannot rely on one of them without doing the appropriate testing," he said.
In response to Dweik's comments, Yeh said that "if a new statistically significant pattern emerges and can be clustered with other like patterns, then we will likely have sensed something new." As an example, "if a variant of a virus comes along and causes an inflammatory response that is uniquely different, our patterns will be different, as well."
"That new 'thing' may not have a name, but we'll know something new can come," Yeh said.
One question for any sensor array, including Vocxi's test, is whether it can differentiate disease states or distinguish between lung cancer and other diseases, Dweik said. Many sensors "work well in the lab because it's a controlled environment," but because the breath has byproducts of the gut microbiome, medications someone is taking, and other compounds, real-world use of sensors for breath testing can be more complicated, he said.
During the COVID-19 pandemic, many breath test developers set their sights on testing for infectious diseases, such as COVID-19 or other respiratory illnesses. Frisco, Texas-based InspectIR received the first Emergency Use Authorization from the US Food and Drug Administration for a breath-based COVID-19 test that used mass spectrometry to detect five VOCs created by a viral infection. Meanwhile, Finnish firm Deep Sensing Algorithms received CE marking in 2022 for a nanosensor-based COVID-19 breath test, and Welsh company Imspex nabbed CE marking the same year for its COVID-19 breath test using gas chromatography and ion mobility spectrometry.
Other firms, such as Avisa Diagnostics and Breathonix also developed SARS-CoV-2 breath tests during the height of the pandemic.
Before the pandemic, however, cancer was the "most popular area of research" for breath testing, Dweik said. Owlstone Medical has long been developing breath tests for lung cancer and other cancers on its Breath Biopsy platform, while researchers in the Netherlands reported in 2019 that their electronic nose test was highly accurate when identifying whether patients with lung cancer responded to certain anti-PD-1 therapies.
Breath testing firm Blu Biotech is also developing tests for lung cancer, colorectal cancer, sepsis, and other diseases using gas chromatography.
Dweik noted that Vocxi's machine learning approach is a "novel way of looking at things," but said that developers should always use caution "to make sure we know what we're looking at and doing it right."