About three-quarters of children hospitalised for fever don’t receive a diagnosis. Current diagnostic tools, which look for pathogens, are slow and sometimes unreliable, says Myrsini Kaforou at Imperial College London. Genetic tests are a promising alternative as certain genes switch on or off in response to disease.
Kaforou and her colleagues gene expression in blood samples from 1212 children who were between a few weeks to 18 years old. All had been diagnosed with one of 18 infectious or inflammatory diseases that cause fever.
The researchers used a machine learning model to analyse this data, and identified 161 genes that correlate with diseases across six categories: bacterial infections, viral infections, inflammatory diseases, malaria, tuberculosis or Kawasaki disease. The researchers validated the model in a separate group of 411 children with fever, and assessed it using a statistical measure that estimates accuracy on a scale of 0 to 1. Across the disease categories, the team found that the model had a score of 0.89 to 1 – generally, a score of 0.8 or higher indicates excellent accuracy.
The finding could speed up diagnoses, ensure appropriate treatment and reduce unnecessary antibiotic use, which contributes to antibiotic resistance, says Kaforou.
“This is a meaningful step in the right direction for precision medicine in infectious and inflammatory diseases in children,” says Christopher Woods at Duke University in North Carolina. However, validation in larger data sets is necessary before it can be used in clinical settings, he says.
“Obviously, this is not a full list of every single infectious and inflammatory disease out there,” says Kaforou. “At the moment, we are recruiting more patients and generating more data to identify a signature that covers even more infectious and inflammation conditions.”