Statistical analysis

Hypergeometric test was used to explore the overlap between the genes in the IgSF-related modules and the top mutated genes in the sub-network for breast cancer. We also studied the enrichment of breast cancer driver genes in IgSF by using this method. Univariate and multivariate analyses were performed using Cox proportional hazards regression model to determine whether the IgSF-related prognostic module was independent of other clinical variables, and adjusted for ER, PR, HER2, age, stage and grade. Hazard ratio (HR) and 95% confidence intervals (CI) were estimated by Cox proportional hazards regression model. To verify if the modules we identified are associated with patient survival, we determined the regression coefficient of every gene in the module related to patient survival using the gene expression data. The classifier was built as a linear combination of the gene expression values of select immune-related genes with the standardized Cox regression coefficient as the weight. A risk score formula for each patient was established by including the expression values of each selected gene, weighed by their estimated regression coefficients in the multivariate Cox regression analysis [42]. Finally, the patients were divided into high- and low-risk groups using the median of the risk score as the threshold. The patients with high-risk scores were classified as poor outcomes. Kaplan-Meier survival plots and log-rank tests by R package “survival” were used to assess the differences in overall survival (OS) time between the high- and low-risk patients. Bioinformatic analysis was performed with R 3.0.0 statistical software.