Fig 1: EMT-ExMap captures topological relation between layers, transition points and stage-specific molecular intricacies of EMT.a Since all datasets have the same set of ten time points, they were first converted to configuration (i.e., similarity) matrices which were then used to compute partial RV coefficients (=matrix correlations) between them (as implemented in iTOP package in R) (Source Data file – Fig. 3a). b The matrix correlations in ‘a’ were used to summarize the topology of interactions between the datasets using the PC algorithm (as implemented in iTOP package in R). c MCIA plot. Different shapes represent the datasets which are connected by lines whose lengths are proportional to their divergence from a common reference point which aims to maximize covariance between the datasets. Only the top 4 highest variation layers (Supplementary Fig. 2c) are being displayed for visual simplicity. d An integrative phylogenetic neighbor-joining tree constructed from all layers except MRNA and METABOL. Since we had protein data, we deemed the transcript measurements redundant. Because METABOL followed a distinct kinetics than other datasets (Fig. 3b), it was treated separately in Fig. 5. Please see the “Methods” section ‘Self-Organizing Maps (SOMs) & Neighbor Joining’ for details. e Top panel. SOM portraits. The color gradient indicates the over- or under-expression of metagenes at each time point compared to the mean expression level of the metagene in the pool of all time points: red=high, yellow/green=intermediate and blue=low. Middle panel. Representative signature molecules that appeared among the highest-ranking features (top 1%) at each time point and layer. Please see the “Methods” section ‘Self-Organizing Maps (SOMs) & Neighbor Joining’ for details of the SOM analysis. Bottom panel. All genes with adj. p-value = 0.01, logFC = 0.6 in each SOM was used as input to perform pathway enrichment analysis using the pathfindR R package. Parameters for pathfindR: pin_name = “IntAct”, p_val_threshold=0.01, gene_sets = “Reactome”, search_method = “GR”, iterations=10. P-values were obtained using one-sided hypergeometric testing followed by Bonferroni adjustment. Representative enriched pathways are shown. f Widefield immunofluorescence microscopy (IFM) images of MCF10A cells treated with TGFß for 6 days were acquired using MICA imaging system (Leica). Antibodies used, and scale bars are shown in the images. Staining intensities were normalized to DAPI acquired at constant settings. Representative images are shown from two independent experiments. g Imaging system and cells same as in Fig. 3f. Antibodies used, and scale bars are shown in the images. Representative images are shown from two independent experiments.