Overview

• Researchers profiled 9,027 samples from 8,262 individuals, measuring more than 5,400 circulating proteins to build the Human Disease Blood Atlas.
• Adult blood proteomes showed stable, person-specific molecular fingerprints, while protein profiles shifted markedly during puberty before stabilizing in adulthood.
• Pan-disease comparisons revealed that many proposed biomarkers recur across unrelated conditions, enabling separation of shared inflammatory signals from disease-specific changes.
• Some disease groups, including pediatric and liver-related conditions, formed distinct proteomic clusters, whereas common cancers often overlapped in their protein signatures.
• Machine-learning models predicted lung, colorectal, and ovarian cancers from protein profiles, with signals for lung cancer sometimes detectable years before diagnosis, though the authors stress the need for independent validation before clinical use.