Modeling osteosarcoma progression by measuring the connectivity dynamics using an inference of multiple differential modules algorithm

Understanding the dynamic changes in connectivity of molecular pathways is important for determining disease prognosis. Thus, the current study used an inference of multiple differential modules (iMDM) algorithm to identify the connectivity changes of sub‑network to predict the progression of osteosarcoma (OS) based on the microarray data of OS at four Huvos grades. Initially, multiple differential co‑expression networks (M‑DCNs) were constructed, and weight values were assigned for each edge, followed by detection of seed genes in M‑DCNs according to the topological properties. Using these seed gene as a start, an iMDM algorithm was utilized to identify the multiple candidate modules. The statistical significance was determined to select multiple differential modules (M‑DMs) based on the null score distribution of candidate modules generated using randomized networks. Additionally, the significance of Module Connectivity Dynamic Score (MCDS) to quantify the dynamic change of M‑DMs connectivity. Further, DAVID was employed for KEGG pathway enrichment analysis of genes in dynamic modules. In addition to the basal condition, four conditions, OS grade 1‑4, were also included (M=4). In total, 4 DCNs were constructed, and each of them included 2,138 edges and 272 nodes. A total of 13 genes were identified and termed 'seed genes' based on the z‑score distribution of 272 nodes in DCNs. Following the module search, module refinement and statistical significance analysis, a total of four 4‑DMs (modules 1, 2, 3 and 4) were identified. Only one significant 4‑DM (module 3 in the DCNs of grade 1, 2, 3 and 4 OS) with dynamic changes was detected when the MCDS of real 4‑DMs were compared to a null distribution of MCDS of random 4‑DMs. Notably, the genes of the dynamic module (module 3) were enriched in two significant pathway terms, ubiquitin‑mediated proteolysis and ribosome. The seed genes with the highest degrees included protein phosphatase 1 regulatory subunit 12A (PPP1R12A), UTP3, small subunit processome component homolog (UTP3), prostaglandin E synthase 3 (PTGES3). Thus, pathway functions (ubiquitin‑mediated proteolysis and ribosome) and several seed genes (PPP1R12A, UTP3, and PTGES3) in the dynamic module 3 may be associated with the progression of OS and may serve as potential therapeutic targets in OS.