Comparative and Sex-Specific Assessment of Proteomics, Clinical, and Genomic Data on Visceral Adipose Tissue Volume Using Machine Learning in UK Biobank: Advancing Precision Medicine
Zahra Azizi, Trisha Gupte, Pik Fang Kho, Jiayan Zhou, Kevin Nzenkue, Rodrigo Guarischi-Sousa, Daniel J. Panyard, Ming-Li Chen, Fahim Abbasi, Shoa L. Clarke, Joshua W. Knowles, Philip S. Tsao, Themistocles L. AssimesBACKGROUND:
Visceral adiposity is a key driver of insulin resistance, which in turn is a strong risk factor for type 2 diabetes and cardiovascular diseases.
METHODS:
We investigated the overall and sex-specific associations between plasma proteomics and visceral adipose tissue volume, and their intersection with clinical and genomic data to identify critical candidate proteins for visceral adiposity possibly linked to insulin resistance. We analyzed data from UK Biobank participants, using multistep machine learning to establish linear and nonlinear associations.
RESULTS:
The combined model, integrating clinical, proteomic, and polygenic data, accounted for 67% of visceral adipose tissue variance. Our research revealed significant protein interactions with environmental factors and identified sex-specific proteins correlated with visceral adipose tissue variance in both males and females.
CONCLUSIONS:
The findings suggest that a small number of plasma proteins explain a significant portion of the variance of visceral adipose tissue beyond nonproteomic measures, offering a potential single-source, personalized measurement of visceral adiposity that is easily accessible and may facilitate improved prediction and understanding of the biology of insulin resistance.