To verify the tissue/disease-relevance of the AI networks, GO term enrichment analysis and hub gene detection and evaluation were performed. The networks analyzed were; the Gastrointestinal – AI - Ulcerative Colitis (UC) network, the Breast – AI - Breast cancer/tissue network, and the Blood – AI - General network. First, we used the Louvain algorithm (resolution = 1.0, minimum module size = 10 genes) to detect important modules in the network. Top hub genes of each module were detected, ranked by their degree of centrality. GO term enrichment analysis of the genes within each module was performed using enrichR (1), and the results were evaluated for disease and/or tissue relevant biology.

In the AI UC network, eight modules were identified. Pathway enrichment results of some of these modules showed key pathogenic functions, such as; Ubiquitin-Dependent Protein Catabolic Process, Immune System, and Neutrophil Degranulation (Table 1 and 2). This indicates that vital UC disease biology can be captured based on the AI networks.



Table 1. AI UC Network, Module 1, Top 10 Significantly Enriched REACTOME Terms.



Table 2. AI UC Network, Module 1, Top 10 Significantly Enriched GO Biological Process (BP) Terms.

In the AI Breast Cancer network, nine modules were identified. Pathway analysis of the largest of these modules showed functions such as Cilium Assembly and Cilium Organization (Table 3). These functions, which are frequently inhibited or lost in breast cancer, can contribute to cancer progression by promoting cell proliferation and metastasis. Another module showed important immune-/cancer-related functions, such as Inflammatory Response, Cellular Response to Cytokine Stimulus, and Antigen Receptor-Mediated Signaling Pathway (Table 4). These pathways are highly relevant to the tumor immune microenvironment and suggest this module may be involved in immune cell infiltration and inflammatory signaling in breast cancer. Furthermore, one module showed enrichment of the Cell Cycle pathway, which is a hallmark of cancer. One of the hub genes within this module, BUB1, is also a well-known driver gene for breast cancer. These results indicate important cancer-relevant processes being captured by the AI Breast Cancer network.



Table 3. AI Breast Cancer Network, Module 1, Top 10 Significant Enrichment GO Biological Process (BP) Terms.



Table 4. AI Breast Cancer Network, Module 4, Top 10 Significant Enrichment GO Biological Process (BP) Terms.

In the AI Blood network, 10 modules were identified. This network includes a diverse collection of blood diseases including myeloid leukemia, pneumonia, and COVID-19. Thus, in the enrichment results, we observed functions involved in regulating viral replication and immune responses, such as mRNA Splicing, Via Spliceosome, Innate Immune System and Neutrophil Degranulation (Table 5 and 6). One module additionally showed important hub genes, MELK, BUB1, CHEK1. These are critical cell cycle checkpoints, frequently dysregulated in myeloid leukemia and other blood cancers, where they contribute to uncontrolled cell proliferation and genomic instability.



Table 5. AI Blood Network, Module 0, Top 10 Significant Enrichment GO Biological Process (BP) Terms.



Table 6. AI Blood Network, Module 0, Top 10 Significant Enrichment REACTOME Terms.