Additional file 8: Table S3. of Extracting a low-dimensional description of multiple gene expression datasets reveals a potential driver for tumor-associated stroma in ovarian cancer

The properties of the modules learned by INSPIRE. Column A: ID of the module. Column B: The number of the neighbor modules in the learned module network. Column C: IDs of the neighbor modules. Column D: The number of genes the module contains. Column E: The names of the genes the module contains. Column F: The number of MSigDB C2 categories for which the module is significantly enriched (Bonferroni-corrected p <0.05) based on a Fisher’s exact test. Column G: The MSigDB C2 categories for which the module is significantly enriched and the corresponding Bonferroni-corrected − log10 p values. The enriched categories are ordered from most significant to least significant. Columns H–L: The p value from the Pearson’s correlation of the feature corresponding to the module with six phenotypes: percent stroma (column H), percent tumor (column I), vessel formation and abundance (column J), stroma type (column K), invasion pattern (column L), residual tumor (column M), and survival (column N). As in Fig. 5a, we highlight modules 5, 6, 53, 54, 60, 78, and 81 that are significantly correlated with at least three of the six phenotypes in red. We also highlight module 30 since it is the only module that has a significant correlation with the vessel formation phenotype. (XLSX 258 kb)