Skip to main content
Premium Trial:

Request an Annual Quote

Predigen Diagnostics Developing Gene Expression Platform to Differentiate Bacterial from Viral Infections

Premium

NEW YORK (GenomeWeb) – Duke University spinout Predigen Diagnostics is developing a multiplex platform that uses host gene expression signatures as a basis for the prognostic, diagnostic, and therapeutic monitoring systems it is developing.

If all goes well, it expects to have a first product in regulatory review by the US Food and Drug Administration sometime in 2022.

One of its first products will likely be a test that helps clinicians do a better job of differentiating bacterial from viral infections, which could help mitigate some antimicrobial resistance concerns, Ephraim Tsalik, a cofounder of Predigen and associate professor of medicine at Duke, said in an interview.

The firm is developing its multimarker platform for a range of medical indications, including infectious diseases, cardiovascular disease, autoimmune and inflammatory diseases, drug response, and cancers.

Predigen's research and development work uses systems biology and machine learning to define the host response to health and disease. Changes in gene expression drive the body's overall response, producing messenger RNA that become the building blocks for proteins sent to counteract the infection.

According to Predigen, its most advanced signatures are for infectious disease diagnostics, including the presymptomatic detection of viral and bacterial infections and the ability to accurately discriminate viral from bacterial infections to enable appropriate prescription of antiviral and antibiotic medications.

In addition to differentiating viral from bacterial infections, a test being developed by the firm could indicate whether there is no infection in a patient or if there is a co-infection. A separate test is being developed to identify a viral infection up to 48 hours prior to symptoms, allowing for early treatment and faster recovery of infected patients, and reducing the risk of an infection spreading among patients.

"Respiratory infections and symptoms are the most common reason for acute-care visits to providers in the US and globally," Tsalik said.

Tsalik is also the principal investigator in an ongoing study to determine whether low levels of procalcitonin can reliably reveal whether a person’s lower respiratory tract infection will improve with antibiotics.

In the study, the National Institute of Allergy and Infectious Diseases-supported researchers with the Antibacterial Resistance Leadership Group are collaborating with researchers and medical experts from BioMérieux.

The US Food and Drug Administration has cleared several diagnostic assays that measure procalcitonin as a host-response biomarker and clinicians are using them in the diagnosis and treatment of sepsis and to help them differentiate between bacterial and viral infections.

However, a study published last year in the New England Journal of Medicine said that the use of procalcitonin testing for lower respiratory tract infections did little to reduce antibiotic use in hospitals.

Looking into the details of the study, a lack of the adherence by some physicians with an algorithm recommending withholding or prescribing antibiotics was seen, Tsalik said.

A lack of clinician confidence in existing diagnostic tools to help them make better decisions about when to prescribe antibiotics is among the reasons companies such as Predigen and Burlingame, California-based Inflammatix have turned to gene expression to help mitigate that dilemma.

In Predigen's platform in development, PCT did not make the cut, "and we've shown, as well as many other groups, that gene expression is a much more robust way to differentiate these types of infections," Tsalik said.

Tim Sweeney, cofounder and CEO of Inflammatix, who is not involved in developing the Predigen test, but has worked with Tsalik in the past, said in an interview, "It is now resoundingly clear to me that physicians are waking up to the possibility of host response answering some of their key clinical questions."

Based on clinical results that he is seeing, Sweeney said a large market need exists "that supports a substantial expansion of host response testing, certainly in the emergency department and in the intensive care unit and eventually in many other care settings, including for inpatients and outpatients wherever they show up with suspected acute infections."

Within the host-response diagnostics space generally and gene expression in particular, diagnostic industry participants continue to see people trying to develop a single biomarker diagnostic platform, Sweeney said. "First, I don't think that any of the data support a single biomarker being better than multiple markers. Second, if you take a longer view, none of those [single-marker] tests can be improved over time."

That drawback is eliminated by combining multimarker testing with machine learning, he said.

With use of a machine learning-based multimarker test on an increasing number of patients, "algorithms can be improved over time, surpassing what any individual biomarker can supply to a physician in terms of clinical knowledge," Sweeney said.

Sweeney said that Inflammatix's first diagnostic tests based on gene expression, including its HostDx Sepsis assay with 29 mRNAs, are probably about two years from commercialization.

One of the challenges with developing gene expression tests and using machine learning algorithms for analysis is that they sometimes work well and are highly accurate when the algorithm is being trained or developed. However, its performance doesn't always translate when the platform is applied to a broad patient population. Inflammatix has shown that it has overcome that challenge and internally validated that a fixed-weight algorithm operates at high performance and is highly likely to also be externally validated as it moves closer to clinical applications, Sweeney said.

Meantime, Predigen has been working with a few diagnostic companies with a view to developing a near point-of-care test that would take a patient sample, most likely blood, and provide an analysis of a patient's immune profile in under an hour, a timeframe needed to make fast clinical decisions.

The firm is currently enrolling about 400 patients in a clinical trial at 10 different clinical sites in the US. It is banking samples in anticipation of advancing development of a diagnostic platform in collaboration with Qvella that it anticipates validating using the clinical samples.

"Qvella has more of a focus on sepsis, and so what we've done with them [is] … use their platform, which is a 45-minute sample-to-answer test, to measure host gene expression signatures" for diagnosing viral infections, Tsalik said.

At the European Congress of Clinical Microbiology and Infectious Diseases in Madrid, he and his colleagues presented a poster describing a fast method to profile gene expression of the host response to infection and inflammation for patient stratification. They used the Qvella Fast-HR process, a sample treatment method capable of performing transcriptomic profiling of whole-blood leukocytes and reported 98.5 percent accuracy using sparse logistic regression classification.

The researchers collected whole blood from 68 people in Beckton Dickinson PAXgene Blood RNA tubes and performed the Qvella Fast-HR process to release stabilized mRNA in an RT-PCR assay-ready medium. They detected target and reference genes by multiplexed real-time RT-PCR.

The team reported that the process achieved accurate initial detection and discrimination of viral vs. non-viral illnesses in less than an hour.

In 2016, Tsalik and his colleagues published the results of a study in Science Translational Medicine in which they measured peripheral whole-blood gene expression from 273 subjects with community-onset acute respiratory infection or noninfectious illness, as well as 44 healthy controls. They applied sparse logistic regression to develop classifiers for bacterial acute respiratory infection, using 71 probes; viral acute respiratory infection, using 33 probes; or a noninfectious cause of illness, using 26 probes. They reported overall accuracy of 87 percent, which they said was more accurate than procalcitonin at 78 percent and three published classifiers of bacterial versus viral infection with accuracies ranging from 78 percent to 83 percent. 

Predigen is currently engaged with four other undisclosed companies "using different approaches that will allow us to get these signatures into multiple different markets," Tsalik said, adding that the firm is interested in developing the platform for use in hospitals, doctors' clinics, and potentially to sell directly to consumers.

The firm plans to eventually apply for regulatory clearance with the FDA and has already engaged in discussion with the agency. "The FDA has a clear recognition of the fact that this could be a very useful tool and has provided useful and informative guidance in terms of" the firm's path toward regulatory clearance, he said.

Predigen's goal is to make the tests "as affordable as possible," he noted. "Because we're working with multiple different approaches, in some cases the unit may be priced in the range of hundreds of dollars, or less, as opposed to tens of thousands of dollars" for molecular multiplex tests, he said. Single test consumables would probably be priced at dollars to tens of dollars, though it is somewhat too soon to discuss price points, he noted.