Fig. 7

Classifier accuracy across cell types. 46 MIEs are plotted by internal accuracy score for models trained using MCF7 derived data (X axis) and for models trained using PC3 derived data (Y axis). For both cell types, internal accuracy scores are computed based on the SVM Linear (SVM_L) algorithm using landmark genes as the feature type. The dashed identity line is shown in green (slope = 1, intercept = 0). The best fit line is shown in black (slope = 0.55, intercept = 0.29). The correlation coefficient R and corresponding t-test p-value are shown at the top of the plot. Points are shaded by the disparity in expression of the associated MIE target gene(s) between cell lines, based on the log2 fold change in NX values from the Human Protein Atlas. Red and blue points correspond to higher relative expression in MCF7 and PC3 cell lines, respectively. For visualization purposes, values greater than 4 and less than − 4 were capped at 4 and − 4, respectively