NEW YORK (GenomeWeb) – A University of Edinburgh-led team derived an RNA signature that can distinguish newborns suffering from sepsis from those without infection.
As they reported in Nature Communications today, researchers led by Edinburgh's Peter Ghazal isolated RNA from infant blood samples to search for RNAs that were differentially expressed between healthy babies and ones with an infection. From this, they devised a 52-gene classifier that could predict bacterial infection.
"Infection is responsible for a significant proportion of neonatal deaths worldwide, and also increases the risk of long-term disability in survivors," first author Claire Smith, a neonatologist at the Royal Infirmary of Edinburgh, said in a statement. "There is a pressing clinical need for more accurate and rapid testing for neonatal infection than is currently available."
Newborns, the researchers noted, are more susceptible than older people to infection, and sepsis can be difficult to diagnose in them as typical symptoms may not arise. Also, due to their immature immune systems, infants' response to infection may differ from that of adults.
Using a training set consisting of 27 patients who were blood-culture positive for sepsis, one patient with a cytomegalovirus infection, and 35 matched controls, the researchers covered a number of signals using a microarray-based genome-wide transcriptional analysis. Principal components analysis separated these groups into control and infected groups, with the exception of the virally infected case, which the researchers then excluded from further analysis.
After filtering, the Edinburgh team homed in on a set of 824 probes that were differentially expressed. Euclidean distance-based clustering revealed an upregulated group, a downregulated group, and third group that didn't have a clear distinction between cases and controls. Further analysis revealed that many of the probes in the third group are involved in blood development, and those, too, were removed from follow-up analyses.
Even more stringent filtering whittled the probe set down to 52 genes, including both upregulated and downregulated ones. These probes, the researchers reported, belong to three classes of functional pathways: innate immunity, adaptive immunity, and sugar and lipid metabolism.
"Strikingly, we find a highly exacerbated and unbalanced homeostatic systemic innate-immune and metabolic response, which is accompanied by a reduced lymphoid involvement. The strength and magnitude of these responses show the remarkable ability of the neonate to recognize and acutely mount a vigorous, but highly focused response," the researchers said. "The response is not a 'storm' but instead exhibits an altered set-point in the regulatory communication between innate and adaptive immune cells."
Drawing on a number of machine-learning algorithms, the researchers reported that their 52-gene dual network has high sensitivity and specificity.
They additionally tested the classifier on a set of 18 infected and 24 controls culled from the training set sample, but using a different microarray platform that interrogated 48 matched genes. The classifier still correctly assigned all of the samples to the bacterially infected or control groups, the researchers said.
To validate the classifier, the researchers turned to an independent set of 26 samples — 16 infected and 10 controls — that were analyzed using three microarray platforms. Here, too, the researchers reported that all samples were correctly assigned to control of infected groups.
For a group of 30 patients who were suspected of having an infection, but whose blood cultures had come back negative, the classifier assigned 17 of those patients to the infected group. Expert assessment of the patients, meanwhile, indicated that six had an infection, 15 did not, and nine couldn't be categorized.
Ghazal noted in a statement that he and his colleagues are now exploring whether a small drop of blood is enough of a sample on which to use this classifier.
"This work is enabling us to move towards being able to distinguish between babies with true infection who need urgent treatment, and those who are not infected and therefore don't require antibiotics," Smith added.