Extended Data Fig. 4: k-means clustering reveals meaningful longitudinal patterns.
From: Early prediction of preeclampsia in pregnancy with cell-free RNA
a, c, The chosen k clusters (dashed line) comparing a performance metric, the sum of squared distances, and values of k for clustering of DEGs for preeclampsia versus normotensive related to Fig. 2d (a) and DEGs for preeclampsia with versus without severe features related to Fig. 4a (c). b, d, Following permutation of the data matrix prior to k-means clustering, longitudinal changes over gestation are replaced by 2 flat lines for clustering of log2(FC) for preeclampsia versus normotensive (b) and 4 uninformative lines for clustering of log2(FC) for preeclampsia with versus without severe features (d). For b, d, points indicate median per trend and shaded region indicates 95% CI.
