A significant international research effort has revealed that dozens of common diseases leave distinct molecular traces in the proteins circulating through human blood, offering a path towards far more precise diagnostic testing. By analysing how thousands of proteins change with age and illness, the team has created a reference tool that may help clinicians distinguish between harmless biological fluctuations and genuine markers of disease. Rather than focusing on a single condition at a time, the study compares fifty-nine diseases side by side, enabling researchers to separate signals that are shared broadly—such as inflammation—from those that point to specific pathologies.
Published today in Science, the work maps how blood proteins behave across a wide range of severe conditions, including cancer, cardiovascular disorders and autoimmune diseases. The researchers have compiled these data into the Human Disease Blood Atlas, which documents the changing landscape of blood proteins not only during illness but also throughout the normal course of human life. The atlas shows that every person has a distinctive blood protein profile, a kind of molecular fingerprint that shifts through childhood before stabilising in adulthood. In clinical settings, such stable individual baselines might someday be used to flag early changes that precede symptoms, transforming diagnostics from reactive to preventive.
The project’s senior author, Mathias Uhlén, and lead author, María Bueno Álvez, explain that the study employed machine-learning approaches designed to identify molecular signatures that remain reliable outside controlled laboratory testing. By training algorithms on multiple diseases simultaneously, the team demonstrated that many proteins rise and fall across unrelated conditions, particularly those tied to inflammation. These overlapping responses frequently mislead researchers in traditional biomarker studies, which typically compare a disease to a healthy control group without considering how proteins behave across different illnesses.
Uhlén, a professor at KTH Royal Institute of Technology and director of the Human Protein Atlas project, argues that this broader comparative approach is essential. “When diseases are analysed side by side, we can separate universal inflammatory alarm bells from signals that are truly specific to individual disorders,” he says. He emphasises that distinguishing these patterns is critical to developing blood tests that do not misclassify patients or yield misleading results in real clinical settings. Some disease signatures cluster by organ system, such as those related to the liver, whereas others reveal shared pathways that span conditions like cancer, autoimmune illness, and infection.
The study also highlights a significant challenge in modern biomedical research: reproducibility. According to Bueno Álvez, around seventy biomarker papers are published each day, yet many are not reproducible because they test only one condition against controls. Proteins that change in several diseases are frequently misrepresented as unique markers. By exposing these shared features and identifying reliable, disease-specific patterns, the Human Disease Blood Atlas offers a corrective framework and demonstrates how future biomarker discovery can avoid this pitfall.
Among the atlas’s most intriguing findings are early protein changes detected in individuals approaching cancer diagnosis, suggesting that some molecular shifts precede clinical signs by a significant span. This glimpse into pre-symptomatic disease suggests the long-term potential of proteomics for early detection. As the atlas continues to expand, it sets the stage for diagnostic tests that can interpret the complexities of blood proteins with unprecedented accuracy, reducing false alarms while capturing disease at its earliest stages.
More information: María Bueno Álvez et al, A human pan-disease blood atlas of the circulating proteome, Science. DOI: 10.1126/science.adx2678
Journal information: Science Provided by KTH, Royal Institute of Technology
