Whole-Blood RNA Test Predicts Disease Trajectory and Treatment Response

Clinicians often must predict whether acutely ill patients will recover or deteriorate despite limited time and clinical evidence. Earlier prognostic information could improve triage and guide treatment decisions across infectious and chronic diseases. Although blood-based transcriptomic signatures show promise, most require serial sampling or focus on disease classification rather than progression. New findings now demonstrate a whole-blood method that predicts disease course and treatment response from a single sample.

Researchers at Imperial College London have developed VeloCD, a computational approach that analyzes whole-blood RNA to anticipate future clinical states. The method is designed to predict whether patients are likely to improve or worsen and whether they will respond to therapy, using information captured at a single time point. The work appears as a proof-of-concept with multiple external validations.

When the body encounters infection or other illness, cells rapidly alter gene activity as part of a coordinated immune response. These changes can be tracked through messenger RNA (mRNA) molecules circulating in blood, which reflect which genes are being switched on or off. Previous international studies, including DIAMONDS, PERFORM, and EUCLIDS, showed that distinct RNA expression patterns can distinguish bacterial from viral infections in febrile children, raising the possibility of faster and more accurate treatment decisions. Building on this concept, the latest work extends transcriptomic analysis beyond identifying the cause of disease to predicting what will happen next: whether a patient is likely to improve or deteriorate and how they may respond to therapy.

VeloCD adapts RNA velocity—originally devised for single-cell analyses—to whole blood samples. The original RNA velocity approach was developed to predict how individual stem cells would mature into different cell types based on evolving gene-expression patterns. As genes become activated, DNA is transcribed into an initial “raw” RNA transcript containing non-coding introns that must later be removed to produce mature RNA. By comparing the relative abundance of raw versus processed RNA, researchers can infer whether gene activity is increasing or declining. In adapting this concept to blood samples composed of millions of cells, the investigators developed a computational framework capable of tracking the trajectory of illness rather than the fate of individual cells.

Using bioinformatics techniques, the system evaluates whether an individual’s evolving blood-expression profile is becoming more or less similar to patterns associated with outcomes of interest, such as severe illness, mild disease, or self-resolving infection. This directional information can be generated without repeated sampling, potentially enabling earlier insight into patient trajectory from a single blood draw.

To validate the approach, investigators reanalyzed data from several large studies, including the European Union–funded PERFORM cohort of almost 400 febrile children treated at hospitals in nine European countries. Whole-blood RNA sequencing identified more than 2,300 markers associated with mild, moderate, and severe illness; a focused panel of 59 markers enabled VeloCD to predict progression and flag children most likely to deteriorate and require intensive care. In Imperial College London’s human challenge program, blood collected as early as Day 2 after exposure predicted who would later develop influenza or COVID-19, preceding confirmation by polymerase chain reaction (PCR). Additional validations indicated potential to highlight complications of HIV and tuberculosis and to anticipate response to therapy in inflammatory bowel disease after the first treatment dose.

The researchers note that the current approach requires whole-blood RNA sequencing, which remains relatively time-consuming. However, they believe the system could eventually be refined into a smaller targeted panel focused on key markers suitable for routine clinical testing. The long-term aim is a point-of-care prognostic assay capable of stratifying patients according to future risk, helping clinicians determine whether seemingly stable patients may deteriorate and require hospitalization or whether they are likely to recover with outpatient treatment.

The study was published in Nature Communications on May 6, 2026. Collaborators included University College London, University of Cape Town, Queen Mary University of London, the PERFORM consortium, and the Imperial human challenge program. The team has filed a patent on VeloCD and made the tool available on GitHub. They state that further development and validation are needed, with the aim of refining the approach into a clinically deployable prognostic test that could help triage patients and guide care.

“Our approach uses cutting-edge methods to provide a glimpse into a patient’s future, based on how their body is responding to illness at that moment in time,” said Dr. Claire Dunican, Research Associate and Bioinformatician in the Department of Infectious Disease at Imperial College London, who developed, adapted and tested the method, and  is the first and co-corresponding author of the study.

"The patterns of gene expression we see in the blood offer clues as to what is happening. By identifying key patterns, we can essentially predict the trajectory of illness – not just where someone is right now, but where they are going to be in  thenext few hours or days. In practice, this could tell us whether they will get better or deteriorate, and how they might respond to treatment," said Dr. Dunican.

Related Links:
Imperial College London