Fig. 1: PurIST subtype prevalence and probabilities between sequencing methods.

A Confusion matrix comparing classical and basal-like subtype prevalence by whole transcriptome and exome capture with Cohen’s kappa coefficient. B Paired boxplots of basal-like probabilities between whole transcriptome and exome capture. Points with connecting line represent the same sample. Line color is gray if there is no subtype switch and orange if there is a subtype switch between the two methods. One-tailed Wilcoxon Signed Rank test was used to determine whether there was a shift towards higher basal-like probability by exome capture RNA-seq. Small amount of random noise was added in x and y directions (maximum 0.05 in x direction and 0.02 in y direction) for visualization purposes only. C Scatter plot of basal-like probabilities between whole transcriptome and exome capture. D Plot of basal-like probability from whole transcriptome and probability discrepancy (basal-like probability using whole transcriptome minus basal-like probability using exome capture). Color of point outline is subtype by whole transcriptome and point fill is subtype by exome capture. EC-seq, exome capture RNA-seq; WT-seq, whole transcriptome RNA-seq.