NEW YORK (GenomeWeb) – In a study appearing online today in Science Translational Medicine, researchers from Stanford University, the University of Cincinnati, and the Cincinnati Children's Hospital described efforts to develop a straightforward blood expression-based classifier for distinguishing bacterial from viral infections.
Specifically, the team used data from several cohorts to narrow in on seven genes with expression patterns that differed in the blood of individuals with bacterial or viral infections. After verifying the discriminatory power of that seven-gene set in dozens more cohorts, the group combined the expression classifier with a previously reported 11-gene signature for differentiating between infection and inflammation, dubbed the Sepsis MetaScore (SMS), to come up an integrated antibiotics decision model (IADM).
"We can now do a two-stage test: we can first say who has infection, so that's a general inflammation or general host response to infection … [and] we can now find a parsimonious signature of host response that distinguishes between bacteria and viruses," senior author Purvesh Khatri, an immunity, transplantation, and infection research at Stanford University, told GenomeWeb.
In data from more than 1,000 individuals, for example, the researchers found that their IADM could detect some 94 percent of the authentic bacterial infections present, with specificity — the number of positive test results stemming from true bacterial infections — at just shy of 60 percent.
The researchers are pursuing a commercial point-of-care diagnostic device through a Stanford spinout company. They hope to have a device ready for regulatory approval within two to three years, Khatri said. Ideally, the device would be able to distinguish bacterial from viral infections in less than an hour at a price point that is cost-neutral compared to an antibiotic prescription.
"If your test is going to be more expensive than antibiotics, it may be difficult to use it outside the ICU and hospitals," he said. "If you really want this to be used in outpatient clinics, which is where many of the incorrect prescriptions for antibiotics are going to be, you need to make sure it's cost effective."
Although bacterial sepsis becomes more and more deadly the longer it is left untreated, he and his co-authors explained, the over-prescription of antibiotics can lead to drug-resistance and can cause its own morbidity for patients who don't actually have a bacterial infection.
"The rate of inappropriate antibiotic prescriptions in the hospital setting is estimated at 30 to 50 [percent] and would be decreased by improved diagnostics," they wrote, noting that "[t]here is currently no gold standard point-of-care diagnostic that can broadly determine the presence and type of infection."
Several teams have taken a host blood gene expression-based approach to addressing this need. In a study published in Science Translational Medicine earlier this year, for example, researchers from Duke University and elsewhere identified gene expression-based classifiers for identifying individuals without infections, with bacterial infections, or with viral infections.
Still, Khatri and his colleagues argued that prior studies — including their own 2015 paper in Immunity outlining a 396-gene signature for viral infection — require testing at too many genes to be feasible for routine clinical testing.
The SMS, originally tested in trauma patients with systemic inflammation to discern between those with infection and those without, could not distinguish between bacterial and viral infections, he explained.
But results from the Immunity study hinted that it should be possible to pick up blood expression-based differences in patients infected by bacteria or viruses.
To explore this further, the researchers scrutinized available expression databases such as the National Institute of Health's Gene Expression Omnibus or the European Bioinformatics Institute ArrayExpress, focusing on expression patterns associated with the presence of bacteria, viruses, infection, sepsis, fever, or other conditions.
Based on expression data for samples from 426 individuals with viral or bacterial infections from eight cohorts, the team uncovered 72 genes that appeared to be differentially expressed depending on the cause of infection.
The researchers narrowed in on seven genes from that set, which were validated in another 30 independent cohorts representing nearly 1,300 individuals and with blood samples from 96 pediatric patients at Cincinnati Children's Hospital who had systemic inflammatory response syndrome, bacterial sepsis, or viral sepsis.
Results in the latter group were validated using NanoString Technologies' nCounter digital multiplex gene quantitation assays. Though that approach can accurately assess the expression of multiple genes, the study's authors noted, it typically takes four to six hours to run each assay, which may not be rapid enough when diagnosing infections in the clinic.
"[A]lthough the assay confirms that our gene set is robust in targeted measurements, further work will be needed to improve the turnaround time," they wrote.
The IADM — which relied on both the SMS and new seven-gene signature — detected bacterial infections with 94 percent specificity and 59.8 percent specificity in follow-up testing on 1,057 individuals with bacterial infection, viral infection, or other types of inflammation.
Khatri noted that it might be more practical to focus on the seven-gene signature alone in clinical settings where systemic inflammatory response syndrome can be ruled out, such as an outpatient clinic.
Along with efforts to continue developing a clinical diagnostic device, the researchers are involved in clinical trials of the seven-gene signature using a standardized platform and in conjunction with current standard-of-care approaches for diagnosing infections from other types of clinical data.
Future studies are needed to optimize the turnaround time and to determine whether such signatures can be used for stratifying infection patients to find those at greatest mortality risk.
The team is also interested in exploring the biological functions of genes in the SMS and IADM signatures, since many have not been linked to pathogen immune response in the past. And by learning more about the genes that are active during infection, it hopes to find potential avenues for developing broad spectrum, host-directed antiviral drugs.