Abstract P273: Sex-Stratified Machine Learning Analysis of Proteomics, Phenotypic, and Genomic Influences on Visceral Adipose Tissue Volume in the UK Biobank

Abstract

Introduction: Adiposity distribution plays an important role in insulin resistance (IR); sex differences in body composition can influence the risk of cardiometabolic disease (CMD). We aimed to assess sex differences in the intersection of proteomic, genomic and phenotypic factors related to visceral adipose tissue (VAT) and IR. Methods: We examined UK biobank participants with proteomic data generated using the Olink proximity extension antibody assay (PEA). Association between proteomics, genomics and phenotypes with VAT volume was evaluated. To assess the complex relationships between features, a multistep approach was employed. Following evaluation of each feature types, the combined effects of proteomics, genomics and phenotypes were assessed. Then, top 30 features from the final model were examined (Figure). R2 was tested as an indicator of the variance of VAT explained by the features in each model and SHAP values were reported. Results: We analyzed a total of 5342 participants (52.2% female, mean age 54.8±7.9 years). Our analysis revealed significant variation in plasma proteins when added to phenotypes and genomic data in both sexes. Polygenic scores, phenotypic factors, proteomics, and the combined model explained 4%, 55%, 57%, and 60% of the variation in VAT in women, respectively. In men, these values were lower at 2%, 49%, 51%, and 56%, respectively. While there were common proteins between men and women, several sex-specific proteins were noted in women (e.g., FGF21, LDLR, INHBC, CES1, WFIKKN2, CLEC4A, CDHR2, PRSS8, TNXB, ERBB2, F9, THBS2, CCL16) and men (e.g., CA14, ADM, PON3, STC1, MEGF10, ASGR1, REN, FURIN, OXT, PSPN, HSP90B1, CTSD, SERPINB5, CNTN3, CTSB, WFDC12, and ADGRG1). Conclusion: Our study demonstrated the presence of significant sex differences in the plasma proteomics associated with VAT. These findings may provide insights into sex differences in protein expression related to CMD and enable an individualized measurement of adiposity as one means of better understanding the biology of IR.