Discover how mapping plasma proteins to diseases in 53,000 adults is transforming diagnostics, advancing treatments, and paving the way for personalized healthcare solutions.
Study: Atlas of the plasma proteome in health and disease in 53,026 adults. Image Credit: angellodeco / Shutterstock
In a recent study published in the journal Cell, researchers unveiled an interactive atlas of the human plasma proteome in health and disease, providing a valuable resource for advancing precision medicine.
With the global population rapidly increasing and aging, there is a growing demand to improve health and mitigate disease burden. Proteins are the biological effectors for environmental and genetic risk of diseases; they directly reflect the pathophysiological changes and biological processes in humans. Exploring the relationship between proteins and disease can help characterize the biological signatures of disease and health states.
Advances in high-throughput proteomics have facilitated a mechanistic understanding of diseases, risk prediction, biomarker identification, and early detection of adverse drug effects. However, most proteomics studies have focused on a limited number of disease outcomes. Further, while these studies have identified some disease-specific proteomic changes, there is a lack of a comprehensive proteome-phenome atlas to uncover shared biological mechanisms across diseases.
The Study and Findings
The present study presented a large-scale and comprehensive atlas of human proteome-phenome associations. Researchers systematically mapped 2,920 plasma proteins to 1,706 prevalent and incident disease endpoints and 986 health-related traits in 53,026 individuals. This represents one of the most extensive proteomics investigations to date.
Prevalent and incident diseases were defined as those occurring before and after the baseline visits, respectively; blood samples and clinical data were collected during these visits. First, the researchers investigated the relationship between circulating levels of 2,920 proteins and incident and prevalent diseases. They identified 168,100 significant protein-disease pairs, comprising 107,158 protein-incident disease and 60,942 protein-prevalent disease associations.
The top associations were observed in incident genitourinary diseases, which comprised both previously reported and unreported protein biomarkers. Next, proteins were ranked (based on p values), and the number of diseases in which each protein had a first-place ranking was calculated. Six of the top 10 proteins with the most first-place rankings were common to both incident and prevalent diseases.
Notably, 27 proteins exhibited divergent effects on prevalent and incident diseases. For instance, klotho beta (KLB), ADP-ribosyl-transferase 3 (ART3), and desmoglein 2 (DSG2) protein levels were elevated in patients with prevalent type 2 diabetes (T2D); nevertheless, these proteins were identified as protective factors for incident T2D risk. This suggests that these proteins may play opposing roles during different disease stages.
Additionally, the researchers explored associations between proteins and health-related traits. They identified 554,488 significant protein-trait associations, encompassing 782 traits and 2,702 proteins. With most proteins showing multi-phenotypic associations, the team focused on pleiotropic proteins, given their potential as clinical targets. They identified 649 proteins linked to over 50 incident diseases and 434 proteins linked to over 50 prevalent diseases.
Growth differentiation factor 15 (GDF15) showed the most associations with diseases (397 incidents and 205 prevalent diseases). Among protein-trait pairs, 365 proteins showed over 300 significant associations, with GDF15 ranking second for protein-trait associations. Additionally, they identified specific proteins that had a favorable effect on the trait and a protective association with disease.
For instance, insulin-like growth factor binding protein 2 (IGFBP2) correlated with reduced T2D risk and lower alanine aminotransferase levels. Functional enrichment analyses revealed that immune system-related pathways, particularly tumor necrosis factor (TNF) signaling, were the most frequently enriched pathways across diseases. Protein metabolism pathways were also significantly involved in many conditions.
Predictive and Diagnostic Potential
Next, the team investigated the predictive and diagnostic value of proteins by modeling disease risk for each endpoint. The protein-based model achieved high areas under the curve (AUC), exceeding 0.8 for 92 diseases and demonstrated excellent predictions (AUC > 0.9) for nine diseases. It was also significantly more accurate than the demographics-based model in predicting 361 diseases and diagnosing 218 diseases. Integrating proteins with demographics substantially increased predictive accuracy for 417 diseases.
Causal Relationships and Drug Discovery
Further, a Mendelian randomization (MR) analysis was performed to investigate whether proteins played causal roles or were a consequence of disease. The team identified 474 potential causal protein-disease pairs and 4,014 pairs where protein changes were likely consequences of disease. Finally, they explored the disease-linked proteins for promising drug targets.
Among protein-incident and protein-prevalent disease pairs, 38 and 54 pairs had clinical trials or approved drugs, respectively, and 37 repurposing opportunities were identified for established drugs. Safety analyses of potential targets revealed 10 targets with the lowest risk, 26 with potential risks, and 76 with probable risks.
Key Innovations and Conclusions
The study’s most notable contribution is the creation of an open-access proteome-phenome resource, allowing researchers to explore detailed protein-disease and protein-trait associations, enriched biological pathways, and diagnostic/predictive models. This interactive tool is expected to accelerate research in precision medicine.
Together, this extensive plasma proteomics investigation on health and disease phenotypes identified 168,100 protein-disease and 554,488 protein-trait associations. The study demonstrated the superiority of plasma protein-based models for disease prediction and diagnosis over traditional demographic models. Moreover, 474 potential causal proteins were identified, providing promising therapeutic targets.
Limitations
The authors acknowledge several limitations, such as the reliance on plasma samples and the limited diversity of the study cohort, which was predominantly white European. Future studies should validate findings in more diverse populations and explore tissue-specific proteomic data to gain deeper insights into disease pathogenesis.
X-chromosome study reveals hidden genetic links to Alzheimer’s disease