Improved Prediction Accuracy for Late-Onset Preeclampsia Using cfRNA Profiles: A Comparative Study of Marker Selection Strategies

Background: Late-onset preeclampsia (LO-PE) poses substantial risks to maternal–fetal health yet remains more challenging to predict than early-onset preeclampsia (PE), partly due to its stronger association with maternal factors such as obesity and chronic hyper-tension. Methods: We leveraged cell-free RNA (cfRNA) sequencing of maternal plasma in 48 samples—comprising early-onset PE, late-onset PE, and corresponding control groups—to identify LO-PE–specific biomarkers. Differential expression analyses and elastic net regression were used to extract LO-PE gene signatures, with solution paths guiding the selection of the most predictive features. Results: Incorporating these LO-PE signatures into predictive models yielded area under the receiver operating characteristic curve (AUC) values of up to 0.88–1.00, surpassing baseline models that plateaued around 0.69. Pathway enrichment indicated that immune and metabolic processes—including Allograft Rejection and Estrogen Response—were strongly implicated, highlighting genes such as HLA-G, IL17RB, and KLRC4 as potential biomarkers. Combining early- and late-onset signatures in a single model introduced performance trade-offs, empha-sizing the fundamental pathophysiological differences between these sub-types. Conclusions: These findings suggest cfRNA-seq–based signatures can substantially enhance LO-PE screening. Nevertheless, larger cohorts and multi-omics integration will be crucial to establish robust, clinically actionable risk stratification.