Clustering breast cancer cell lines into groups of similar morphology. (A) Heat map of Euclidean distance between cell lines for shape features to illustrate clusters arising from k-means method. The colored lines on the bottom show the assigned cluster and the cadherin expression and assigned canonical cancer subtype. (B) Dot plot showing the enrichment of gene expression modules in the different cell line clusters. Along the y-axis are the names of the clusters, faceted by whether they are included in the PCSF-derived regulatory network on the bottom and whether they are correlated with cell-shape variables, but not included in the network on the top. The x-axis shows the cell-shape clusters, with letters corresponding to the groups in A. (A) A heterogeneous mix of breast cancer subtypes, (B) luminal-like cell lines, (C) basal-like cell lines. Dots are colored based on the normalized enrichment value, with down-regulated modules in blue and the up-regulated modules in yellow. Size corresponds to significance [−log10(P)] with the shape illustrating which changes are significant (adjusted P < 0.01, Benjamini–Hochberg). (C) Images (see Methods) showing morphology of representative cell lines from each respective cluster. Colors indicate labeling with DAPI (blue), anti-p65 (green), and DHE (red).
