(P24-079-24) Harnessing the Plasma Proteome to Assess the Prevalence of Multiple Micronutrient Deficiencies in Early Pregnancy in Rural Bangladesh

Objectives: Multiple micronutrient deficiencies are rarely assessed due to technical challenges and costs, potentially solvable on a single, multiplex, proteomics platform as nutrients require, and covary, with proteins in vivo. We mined plasma proteins of pregnant women in rural Bangladesh to predict micronutrient status and prevalence of deficiencies.

Methods: First trimester, baseline plasma samples of 435 women in the JiVitA-3 micronutrient vs iron-folic acid trial were assessed for biochemical concentrations of 22 biomarkers of vitamins, trace elements, carotenoids, lipids, and inflammation. Slow-Off Rate Modified Aptamer (SOMAmer^® Reagent) binding with fluorescence detection (SomaLogic^® Boulder, CO) yielded relative quantification for 7363 protein targets. Linear regression revealed proteins correlated with nutrients at a false discovery rate (chance-corrected p-value) <0.05, defining nutriproteomes. Forward, stepwise regression identified nutrient predictive proteins, limited by an AIC decrement of > -15, generating models with explained variance (R²). In a randomly selected training set (n=290), nutrient predictors were modeled anew, generating coefficients that were applied to the random test set (n=145). Modeled and biochemical distributions were compared.

Results: Aptamer-bound protein targets varied in number: for retinol (n=377), α-tocopherol (n=1129), cholecalciferol (n=34), zinc (n=303), Se (n=19), Cu (n=2607), thyroglobulin for I (n=3), ferritin for Fe (n=22), cholesterol (n=1237), AGP (n=147), and CRP (n=859). Member proteins comprised transporters, receptors, enzymes, signaling and structural molecules; proteins of nuclear, cytosolic, organelle, and extracellular compartments, among others. Regression models with 1-8 proteins, generating R² values from 0.16 to 0.92, accurately predicted nutrient means (location), while predicting deficiencies requires a high R². Effort to expand the variance to predict prevalence of deficiencies is work in progress

Conclusions: Plasma proteins measured on a single platform might, in the future, predict micronutrient deficiencies although challenges persist with respect to reliability, quantification and instrumentation.

Funding Sources: Johns Hopkins Bloomberg SPH, Kirk Humanitarian and Bill & Melinda Gates Foundation