Fig. 3. Results of the analysis.

a Disease Switch Gene Network (DSGN). Schematic representation of disease modules informed by switch genes in the human interactome. Switch gene products (switch proteins) were colored based on the disorder class to which they belong. Gray nodes and the corresponding links are theoretical and represent non-switch proteins hypothetically connected to switch proteins within the interactome. b Generalized measure of the module separation. The function f(x) is the generalized version of module separation defined in Eq. (3) (cf “Methods”). This function approaches its maximum value when the disease modules are significantly well-separated (p-value < 0.05), whereas it approaches its minimum value when the disease modules significantly overlap (p-value < 0.05). The value of α = 0.3 was chosen to have f(x) close to zero for statistically insignificant p-values (i.e., f(x) in [−0.1, 0.1] as highlighted by the red circle). The blue bars represent the frequency of x values, ranging from −1 and 1.6, Supplementary Data for statistically insignificant p-values, ranging from 0.09 and 1 (Supplementary Data 4). c SWIM-based disease dendrogram and symmetrical heatmap. The diseases modules identified by the disease-specific switch proteins in the human interactome are clustered by a complete linkage hierarchical clustering algorithm and by using the separation metric as a distance metric. Heatmap colors refer to the generalized separation metric, increasing from blue to red: shades of blue refer to cognate disease modules (i.e., s <0, p-value < 0.05); shades of red refer to non-cognate disease modules (i.e., s > 0, p-value < 0.05); and shades of yellow refer to uncertain disease modules (i.e., p-value > 0.05). d SWIM-informed Human Disease Network (SHDN). In the SHDN, each node corresponds to a distinct disorder, colored based on the disorder class to which it belongs. Labeled nodes correspond to the 14 diseases analyzed in this study, while unlabeled nodes are artificial and represent other diseases or developmental endotypes to be investigated. The size of each node is proportional to the number of switch genes involved in the corresponding disorder. A link between two diseases occurs if they share a substantial number of switch genes. AD Alzheimer’s disease, BLCA bladder urothelial carcinoma, BRCA invasive breast carcinoma, CHOL cholangiocarcinoma, COAD colon adenocarcinoma, COPD chronic obstructive pulmonary disease, HNSC head and neck squamous cell carcinoma, IC ischemic cardiomyopathies, KIRP kidney renal papillary cell carcinoma, LUAD lung adenocarcinoma, LUSC lung squamous cell carcinoma, NIC non-ischemic cardiomyopathies, PRAD prostate adenocarcinoma, UCEC uterine corpus endometrial carcinoma.