Skip to main content
Premium Trial:

Request an Annual Quote

Stanford Team Discovers Diagnostic Gene Set for Sepsis; Aims to Develop Clinical Assay


NEW YORK (GenomeWeb) – Analyzing existing gene expression datasets, researchers at Stanford University have discovered an 11-gene set that allows them to identify patients with sepsis and distinguish them from patients with systemic inflammation from non-infectious causes. The team is currently conducting clinical studies to confirm their findings, with the goal of developing a rapid clinical sepsis assay.

Results from the study that identified the 11-gene set were published online todayin Science Translational Medicine. "We've certainly shown that there is information in the gene expression space that can aid in diagnostics that's not being used now," Tim Sweeney, the study's first author and a surgery resident at Stanford, told GenomeWeb. "What we're looking forward to is showing that, in combination with current diagnostics, this will improve our ability to diagnose sepsis," he said.

According to the researchers, about 750,000 people in the US die each year of sepsis — a systemic inflammation syndrome that is caused by infection and is difficult to distinguish from sterile systemic inflammatory response syndrome, SIRS, which can result from non-infectious causes such as trauma or surgery. No blood test other than microbiological culture, which can take several days, currently exists to diagnose sepsis accurately, said Sweeney, who is also a postdoctoral fellow in the lab of Purvesh Khatri, the study's senior author and a researcher at the Stanford Institute for Immunity, Transplantation, and Infection.

A quicker test would enable doctors to start sepsis patients on antibiotics sooner, which is important because studies have shown that each hour of delaying treatment increases mortality by up to 8 percent, Sweeney said. Ideally, a diagnostic sepsis test would yield results within an hour. "You really need to be in that under-an-hour space in order to be useful to a clinician and not just interesting to a scientist," he said.

For their study, the researchers re-analyzed 27 publicly available datasets from microarray-based gene expression studies of 22 independent cohorts that were conducted over the last ten years or so and focused on the diagnosis, prognosis, and pathophysiology of sepsis. While some of the studies only analyzed patients at the time of admission to the hospital, others followed them over the course of their hospital stay.

To discover sepsis-specific gene expression, the researchers conducted a time-matched analysis in a subset of the cohorts, meaning they compared patients with non-infectious inflammation and with sepsis at the same time point after hospital admission, using a previously published multicohort gene expression analysis framework. This yielded a set of 82 genes that were differentially expressed between the two patient groups.

Previous studies had compared data from patients at different time points, for example at the time of admission to when they developed sepsis later on, but those often picked up gene expression signatures of recovery rather than of sepsis and confounded the analyses, Khatri said. Only a time-matched analysis allowed the Stanford researchers to find a true signature of infection.

To narrow the 82-gene set down to the most discriminatory genes, the researchers carried out a so-called "greedy forward search," leaving them with a set of 11 genes. In a validation study, they tested this gene set in independent datasets from 15 of the remaining cohorts and showed that it maintained good discriminatory power. In combination with existing criteria, they showed, the gene signature improved the diagnosis of sepsis.

The Stanford team has filed for intellectual property protection of the gene panel, which it hopes to develop into a clinical assay, likely still two to three years away. "The most difficult thing right now is finding which of the platforms that are available to do gene [expression] quantitation quickly is going to be the best fit for us," Sweeney said.

To further validate its results, the team is conducting two clinical trials — one at Stanford, in collaboration with Angela Rogers, and the other Hector Wong at Cincinnati Children's Hospital.

The Stanford trial seeks to recruit 100 patients admitted to the intensive care unit for sepsis or non-infectious inflammation, and will compare the gene signature with typical clinical markers of infection, like vital signs, white blood cell counts, or biomarkers such as procalcitonin, and see how it improves diagnosis of sepsis. It will measure gene expression by quantitative PCR, either using a commercially available platform or technology currently developed by an unnamed new company at Stanford, Sweeney said.

In addition, the researchers will study whether the proteins themselves that are expressed by the genes in their panel may have diagnostic power. Working with a Stanford core facility, the scientists plan to measure the proteins using a quantitative Western blot technology from ProteinSimple, a platform called Peggy Sue. "If the proteins have diagnostic power, it would be much easier to design a clinical assay" than one that relies on gene expression, Sweeney explained.

The Cincinnati collaborators, who have already collected samples from pediatric patients, plan to measure the gene set using NanoString Technologies' platform.    

Several other groups and firms are also working on clinical sepsis assays that rely on gene expression signatures. Immunexpress, for example, said in April that it is finishing a clinical trial of its sepsis assay and plans to submit the test to the US Food and Drug Administration this month. In addition, a number of companies are working on sepsis tests that rely on detecting pathogen DNA in patients.

Sweeney said what sets his team's assay apart from others is that it only uses a small number of genes with high statistical power. "We really get down to just the core set of genes that are most highly predictive of sepsis, as opposed to other companies that may be using dozens or sometimes even hundreds of genes to try to differentiate the two," he said.