Proteomics networks linking diet to cardiometabolic risk factors: the Framingham Heart Study
The American Journal of Clinical Nutrition, 2025
Lee S., Joehanes R., Huan T., Liu C., Levy D., Ma J.
| Disease area | Application area | Sample type | Products |
|---|---|---|---|
CVD Nutritional Science | Pathophysiology | Plasma |  Olink Explore 3072/384 |
Abstract
Background
Proteomics has facilitated the identification of key pathways linking diet to diseases. However, key challenge in high-throughput proteomics is identifying functional units of proteins that act together in biological processes.
Objective
We aimed to identify protein networks associated with diet quality and CVD risk factors.
Methods
We analyzed 740 Framingham Heart Study (FHS) participants (mean age 52 years; 46% female). Weighted gene co-expression network analysis (WGCNA) was applied to construct protein networks (i.e., modules) using 2,651 plasma proteins. We assessed cross-sectional associations of modules with the Dietary Approach to Stop Hypertension (DASH) diet score, and BMI. We examined prospective association of the diet- and BMI-associated modules with incident fatty liver and type 2 diabetes (T2D). Furthermore, we conducted Mendelian randomization (MR) analysis to investigate protein-protein relationships.
Results
There were 39 protein modules identified, and each was assigned an arbitrary color name. Four protein modules were associated with both diet and BMI. For example, a 10-unit higher DASH score was associated with 0.27 standard deviation (SD) lower Darkgrey module eigenvalues (95% CI: 0.12, 0.42; P = 0.0008), and per 0.27 SD lower Darkgrey module was associated with 1.17 kg/m2 lower BMI, 95% CI: 1.07, 1.27; P = 2.4e-88. Furthermore, we found that the Darkgrey module was associated with both incident fatty liver and T2D, and the association with incident fatty liver remained after BMI adjustment (odds ratio 3.22 per SD increase, 95% CI: 1.68, 6.19; P = 0.0005). The Darkgrey module comprises 39 proteins, including eight proteins such as FABP4, LEP, and IL1RN that may drive with the association with diet and BMI. MR analysis revealed three putative causal protein pairs from the Darkgrey module.
Conclusions
Our findings highlight proteomic networks potentially linking diet and CVD risk and demonstrate the usefulness of proteomics for identifying high-risk individuals for future interventions.