Protein–protein interaction network-based evaluation of lung cancer relevance and hub gene identification from in vitro crocidolite exposure signatures

Objectives This study evaluated whether a protein–protein interaction (PPI) network-based approach can link in vitro exposure-induced gene expression signatures to human disease modules. Methods Crocidolite, a well-established lung carcinogen, was used as a model of inhaled dust. Human immortalized type I alveolar epithelial cells were exposed to five doses of crocidolite, and RNA sequencing was performed. 253 asbestos exposure-responsive genes showing dose-dependent responses were defined as a crocidolite module. Lung cancer-related proteins were compiled from OMIM, GWAS Catalog, PheWAS, and ClinVar, yielding a lung cancer module of 90 proteins. The network proximity between the crocidolite module and the disease modules was evaluated using a z-score. Hub genes associated with both crocidolite exposure and lung cancer were identified by Cross-betweenness, Graph Kernel, and network propagation (NP). Method performance was assessed by receiver operating characteristic curve analysis using 11 guideline- and expert-selected lung cancer driver genes. Results The crocidolite module showed significant proximity to the lung cancer module (Zc = −2.95), but not to the thyroid cancer module used as a negative control (Zc = 2.30). NP showed the highest performance for recovering driver genes (AUC = 0.974). NP-prioritized genes were enriched in PI3K–Akt signaling, cancer-related pathways, EGFR/ErbB signaling, apoptosis, and cellular senescence. The top hub genes included PARP1, STAT3, MYC, EGR1, CYP1A1, and SMAD3. Conclusions Integrating in vitro exposure transcriptomics with human disease modules through PPI-based network analysis may enable the computational evaluation of disease relevance and the identification of biologically meaningful hub genes for the assessment of occupational hazards.