Fig. 5

Classification of NEPC samples based on CTC gene expression profiles. A: normalized gene expression data were simplified to binary representation of marker positivity based on thresholds set by means of healthy donor samples, the frequency distribution of the number of positive NE markers in CTC positive samples is shown for NEPC compared to non-NEPC (AVPC, mHSPC, DNPC); B: ROC analysis of the count of positive NE markers as a marker to distinguish between NEPC and mHSPC, C: ROC analysis of the count of positive NE markers as a marker to distinguish between NEPC and AVPC; D: ROC analysis of the random forest classifier trained on normalized gene expression data of NEPC and mHSPC samples from individual patients positive for CTCs; E: results of the leave-one-out cross-validation included in the classifier training; F: variable importance of the individual markers included in the classifier; G: ROC analysis of the random forest classifier trained on normalized gene expression data of NEPC and AVPC samples from individual patients positive for CTCs; H: results of the leave-one-out cross-validation included in the classifier training; I: variable importance of the individual markers included in the classifier