Fig. 4: The predictive performance of various cfRNA models using the internal and external validation datasets. | Nature Communications

Fig. 4: The predictive performance of various cfRNA models using the internal and external validation datasets.

From: Raised Leptin and Pappalysin2 cell-free RNAs are the hallmarks of pregnancies complicated by preeclampsia with fetal growth restriction

Fig. 4: The predictive performance of various cfRNA models using the internal and external validation datasets.The alternative text for this image may have been generated using AI.

A The AUC scores of the 2- to 10-cfRNA models, chosen by Elastic net, using the 12wk, 20wk, 28wkGA and 36wkGA samples from the internal validation cohort. B the AUC scores of two univariable cfRNA models (LEP and PAPPA2 separately) and the best performing multivariable cfRNA models (LEP + PAPPA2) using the discovery and the validation datasets. The best performing multivariable model, i.e., two cfRNA model (LEP + PAPPA2), was chosen by Elastic net as shown in (A). C The AUC scores of three selected cfRNA models validated on the external dataset (14). D Individual Receiver Operating Characteristic (ROC) curves of two univariable LEP and PAPPA2 cfRNA models using the discovery, the validation cohort, and the external dataset (14). In (B, D), the panel coloured in grey background indicates the training dataset (i.e., the 28wkGA samples of the discovery cohort) where the corresponding models were fitted. Therefore, the AUC is from the same dataset where the model was trained, i.e., not cross validated. In Supplementary Fig. 3, the z-scores of LEP and PAPPA2 are plotted by the gestational age and the type of datasets (i.e., internal and external validation dataset) as the same format in (B). In (A–C) the mean AUC and 95% CI are shown.

Back to article page