Fig. 4

R² signature and permutation analysis of IMC data from 46 breast cancer samples. A Imaging mass cytometry example image from a breast cancer sample (HH3 in blue, CD68 in gray, E. cadherin in red and Vimentin in green) and improvement in the predictive performance (variance explained) for all samples when considering multiple views in contrast to a single, intraview (in absolute percentage points). B The relative contribution of each view to the prediction of the expression of the markers. C Distribution of improvement in variance explained when considering multiple views in contrast to a single intraview across all markers and samples with original cell locations and 10 random permutations. The p-value is calculated by a one-sided Wilcoxon rank-sum test. D Distribution of the relative contribution of the intraview, juxtaview, and the paraview to the prediction of the markers across all markers and samples with original cell locations and 10 random permutations. The p-values are calculated by a one-sided Wilcoxon rank-sum test. E First two principal components of the R² signature of the samples colored by grade and clinical subtype, and the importance of the variables of the signature in the principal component analysis. The naming of the variables is in the form Marker_Measure. The measures taken into account are variance explained by the intraview only (intra. R2), total variance explained by the multiview model (multi. R2), and the gain in variance explained (gain. R2)