Early Prediction and Longitudinal Modeling of Preeclampsia from Multiomics

The study developed a machine learning model for early prediction of preeclampsia (in first 16 weeks of pregnancy) and over gestation by analyzing six omics datasets from longitudinal cohorts of normotensive and preeclamptic pregnant women. The model had high accuracy in early pregnancy using urine metabolome and plasma proteome. A model using only nine urine metabolites had the highest accuracy and was validated on an independent cohort. Univariate analysis further confirmed statistical significance of all metabolites and proteins chosen by the prediction models. Longitudinal modeling revealed biological pathways associated with preeclampsia and confirmed known pathological alterations while suggesting novel associations between the immune and proteomic dynamics. The study identified several biological pathways associated with preeclampsia including steroid hormone biosynthesis (both early in pregnancy and over gestation), tryptophan and caffeine metabolisms (over gestation) and arachidonic acid metabolism (in early pregnancy). Integration with clinical variables further improved prediction accuracy of the urine metabolome model. The findings could lead to a simple, early diagnostic test for preeclampsia.