Fig. 1

Co-expression modules and hubs associated with melanoma progression, EMT, and metastasis in the cellular model of melanoma progression. a, b Matrices containing the module-trait relationships for each cell line (a) or for tumor state, progression, EMT, and metastasis (b). Module names are shown on the y-axis, and correlation coefficients are displayed at the top of each row. Module-trait relationships were identified by calculating the Pearson correlation between module eigengenes and sample features classified using binary or sequential values. For assessing the relationship between tumor behavior and specific modules, the cell lines 4C11− and 4C11+ were assigned to the value 1 and the remaining cell lines were assigned to 0. For the EMT relationship, the mesenchymal-like 4C and 4C11− cell lines were assigned to 1 and the remaining epithelial cell lines were assigned to 0. For the metastasis relationship, only samples from the 4C11+ cell line were classified with the value 1. For tumor progression, the melan-a, 4C, 4C11−, and 4C11+ cell line samples were classified with the sequential numbers 1, 2, 3, or 4. Rows are colored based on the correlation sign of each module with the sample traits: red for positive (red) and negative (blue) correlation. The p values for each module are displayed at the bottom of each row within parentheses. Only modules with a correlation coefficient > 0.3 (absolute value) and a p value < 0.1 were considered significantly associated to a sample trait. c–e Subnetworks containing the co-expression interactions between the hubs from modules positively related to tumor progression (c), EMT (d), and metastasis (e). The hub genes from the modules that showed positive correlation with tumor progression (red, blue, light cyan, and pink), EMT (red, salmon, magenta, and turquoise), and metastasis (midnight blue, greenyellow, yellow, blue, and light cyan) were used to build the subnetworks related to these biological traits. Nodes were colored based on their module assignment, and node sizes were adjusted by degree centrality. Hubs were classified as the 10 top-ranked genes in each module based on intramodular connectivity (Kwithin). Networks were built using Cytoscape