Skip to main content
Premium Trial:

Request an Annual Quote

Duke 'Exposome' Project Nears Commercialization, Spawns Startup


NEW YORK (GenomeWeb) – A long-standing research project to establish RNA biomarker signatures indicative of exposure to different types of pathogens has transitioned to a commercialization phase and is homing in on industry partners.

Initially based at Duke University, the project has been spun out into a company called Host Response, with the new firm exclusively licensing the underlying technology and now seeking to identify potential partners in the platform space, according to Chris Woods, a co-founder of the company and professor of medicine, pathology, and global health at Duke.

Woods said the group has an ongoing clinical trial that runs through 2019 to generate data for submission to the US Food and Drug Administration, and it hopes to find a platform collaborator and to submit a product sometime in the next two years. The platform space is "ever-changing," Woods said, with new technologies debuting seemingly every week. "Luckily, we've been in a position to talk to a lot of different companies and develop diverse relationships based on the use-cases we're seeking to employ," he said.

The firm will be weighing variables such as how many targets can be accommodated, how easy a platform is to use, whether it can potentially be CLIA waived, and the costs of the chemistry, disposables, and instrument.  Cost will be particularly important for point-of-care applications in low-resource settings. "We want to democratize these types of approaches because we want to move them out to more austere environments ... such as a Walmart clinic, or a small village in Africa," Woods explained.  

A test to determine whether an infection is caused by a virus could require just a few targets – Woods suggested the eight- to 10-target range is ideal, which opens the door to many different types of platforms. However, to distinguishing bacterial from viral infections, making a gram-positive versus gram-negative call for bacteria, or determining influenza versus non-influenza viral infections would all require a larger number of targets. In theses cases, platforms that can detect "tens, hundreds, and thousands of targets simultaneously become very attractive," Woods said. 

Exposome project

The 'exposome' project — a multidisciplinary effort anchored at the Duke Center for Applied Genomics to identify signatures that parse host responses to viruses, bacteria, and non-infectious conditions — began in 2001, and in 2009 it landed a $20 million Defense Advanced Research Projects Agency (DARPA) grant.

The group subsequently published descriptions of host responses to bacterial infections and influenza, with the latter showing a detectable gene signature as early as 29 hours after exposure and the highest accuracy around 40 hours before clinical symptoms peaked. The group also reported the development of an RT-PCR-based assay of host response to viral infections in 2013.

Three years later the team published a paper in Science Translational Medicine describing an assay incorporating a 71-probe bacterial classifier, 33-probe viral classifier, and a 26-probe noninfectious disease classifier.

In that work, all three classifiers were applied to a set of samples from 102 adults presenting to the emergency department with fever and 41 healthy volunteers, in order to generate three discrete probabilities for each condition. The TaqMan array-based RT-PCR assay was able to correctly classify these patients with a sensitivity of 89 percent and a specificity of 94 percent.

In his CVS presentation, Woods showed some specific cases that elucidated the way the assay works. In one example, a 16-year-old adolescent with fever, cough, sore throat, and muscle pain, who had a positive PCR test for influenza 2009 H1N1, was classed by the test as having a 99 percent probability of viral infection, a 1 percent probability of SIRS, and less than 0.1 percent probability of a bacterial infection.

On the other hand, a 60-year-old man who developed pleuritic chest pain, a fever, and opacities in a CT scan after having a liver transplant, was found to have Staphylococcus aureus bacteremia. His classifier testing yielded a nearly 100 percent chance of bacterial infection, but also a 21 percent probability of SIRS, and almost one percent probability of viral infection.

The non-infectious example was a 62-year-old woman with non-small cell lung cancer who developed labored breathing, hypoxia, and was coughing up blood, and whose CT showed increased size of masslike opacity. Her classifier test showed a 98 percent probability of SIRS, with a 3 percent probability of viral infection and a 0.1 percent probability of bacterial infection.

"This is a paradigm shift, we're talking about sending out reports that have probabilities, and those need to be interpreted," Woods noted. He said his impression has been that the FDA is attracted to the idea of providing tests results that are probabilities, although it is admittedly a "paradigm shift."

Partner pilot data

Woods presented pilot data from collaborations with BioMérieux subsidiary BioFire and with Qvella at the Clinical Virology Symposium last month.

BioFire is a long-term collaborator, Woods said, and the firm has been a key partner in particular on work encompassing a study called Rapid Diagnostics in Categorizing Acute Lung Infections, or RADICAL, which is supported through the National Institute of Allergy and Infectious Diseases-funded Antimicrobial Resistance and Leadership Group.

In the BioFire pilot data, samples from 50 patients with acute respiratory illness were sorted into bacterial infection, viral infection, or non-infectious systemic inflammatory response syndrome (SIRS) using probes for biomarkers in three separate FilmArray pouch-based assay. The preliminary set showed 88 percent overall accuracy. The team is now working on pruning the number of genes and will present more data later this year, Woods said.

Woods also described a collaboration with Qvella that was presented in poster form at the European Congress of Clinical Microbiology and Infectious Diseases. That pilot study aimed to differentiate viral from non-viral illness using a set of 21 genetic targets, comparing mRNA in whole blood leukocytes from 24 healthy controls, 19 viral infections, 15 bacterial infections, and 10 SIRS cases. The assay used Qvella's platform and proprietary electrical sample prep technique called Field Activated Sample Treatment, or FAST, to detect mRNA using multiplex RT-PCR in less than 45 minutes, demonstrating a 99 percent accuracy in differentiating viral from non-viral illness. 

Interestingly, BioFire's parent company, BioMérieux, has invested in Qvella, and the two firms are reportedly exploring collaborative opportunities in complementary markets using the FAST technology.

Woods said BioFire is an attractive option for a partner in developing panel testing because of the firm's proven track record for getting through FDA, global footprint, and detection of around two dozen probes. Although the exposome project pushes that probe limit, "They certainly have the space to accommodate, it just becomes a more complex technology to develop," he said.

Both of these platforms are PCR-based methodologies with extraction, amplification, and detection in a cartridge or pouch microfluidic system. There are other approaches, such as isothermal amplification, that are also attractive in certain circumstances, and Host Response has relationships with companies in these spaces, as well, Woods said.

At CVS, he noted that the next steps will involve bringing a blood-based test to the FDA, which may be a near-patient or point-of-care test, and potentially also CLIA-waivable.

Competing technologies, the stewardship piece, and 'smart tissues'

There are innumerable non-genetic methods that can be used to try to determine the cause of an acute respiratory illness. For example, a simple white blood cell count can indicate an infection, and C-reactive protein can indicate general inflammation. Procalcitonin tests have some success in parsing viral from bacterial infections and there are a number of PCT assays now approved to diagnose and manage sepsis, although PCT tests fare less well in distinguishing infectious from non-infectious SIRS, Woods said.

Multi-analyte technologies may have an advantage in specificity and sensitivity over single-analyte assays, Woods said, but here too there are many methods, both in development and commercially available, including protein-based tests that would tend to have a greater number of platform options than genomic tests.

In terms of transcriptomics, Inflammatix is developing a one-hour test called HostDx Sepsis that is an 18-gene panel to distinguish bacteial, viral, or no infection from a blood sample. And the Immunexpress Septicyte Lab direct-from-blood test has been cleared by the FDA to differentiate sepsis from SIRS. Immunexpress is also partnering with Biocartis to translate the test to the Idylla RT-PCR platform in order to have a more rapid turnaround time, with a planned commercial release in the US by the end of 2019.

These firms are doing great work, Woods asserted, but the Duke team has a slightly different approach. It has developed a carefully-phenotyped biorepository of adjudicated specimens over the past 15 years that includes co-infections and different vulnerable populations, while other firms tend to rely on databases. "That strengthens the value and rapidity with which we're able to assess very complex questions rather than the more straightforward ones" he said. Other tests also focus on a small number of targets and are perhaps more useful in hospital settings for very ill patients.

In addition, the ability to report probabilities may distinguish the Host Response technology from other approaches.

One of the most important uses of the exposome technology might be to prevent inappropriate use of antibiotics. It has been suggested that pathogen identification alone is sometimes insufficient, and doctors may be prescribing antimicrobials to people who test positive for viral infections because they don't have confidence in the results or don't feel that they rule out bacterial co-infections entirely.

"The greater assurance that a clinician has that the patient does not have a bacterial infection, or has some alternative cause for their symptoms, the less likely they are to use antimicrobials inappropriately," Woods said. He suggested that there is a possibility that acute care physicians might be more comfortable with probability-based test result reporting, as such results would both rule in one type of pathogen and rule out the other, or potentially confirm a co-infection or super-infection.

Ultimately for stewardship, "There has to be a healthcare infrastructure that is comfortable with sending people home without antibiotics," Woods said. Although superinfections can occur, "We hope we have a technology that can detect those very early," he added, although behavioral modification and a support system for prescribers will probably always need to be a component of stewardship.

Finally, Woods sees a future where changes in the ability to rapidly communicate and deliver resources lead to more in-home testing with clinical oversight. Recent studies have shown potential peptide signatures of the host response to viral exposure, including work by researchers at Yale as well as by the Duke team and its collaborators.  Woods suggested this could be translated to a paper-based test using different methods, highlighting techniques being developed by the Whitesides group at Harvard during his CVS presentation. Such a test could perhaps even take the form of a "smart tissue" that would change color, with results reported on a cloud-connected system, and a patient could be immediately sent a prescription, if appropriate, or a viral support pack that includes things like chicken soup, pain relievers, and advice to stay home until they get well.