Organoid-line-specific regulatory network analysis identifies targeted-drug response-associated subnetworks in patient-derived colorectal cancer organoids

Translational interpretation of targeted-therapy response in colorectal cancer (CRC) remains challenging because conventional biomarkers incompletely explain sensitive and resistant states, particularly for anti-epidermal growth factor receptor therapy. We tested whether post-treatment-associated, organoid-line-specific regulatory-network changes could add interpretable context to conventional biomarker analyses in patient-derived CRC organoids. We analyzed organoids from 10 patients treated with five targeted agents. Whole-exome sequencing (WES), paired RNA sequencing (RNA-seq) before and 8 h after drug exposure, and organoid-line-specific regulatory-network analysis were used to compare conventional genomic and expression-based readouts with network-level response features. Selected network-derived genes were examined by quantitative polymerase chain reaction (qPCR) follow-up. To address the small cohort size, empirical response categories, and network-threshold dependence, we also performed internal robustness analyses using ΔECv threshold sensitivity, full-cohort support mapping across response groups, and exact same-size subset permutation. Mutation, copy-number alteration, expression-only clustering, and differential-expression analyses provided only partial stratification. Genomic features were directionally informative in selected settings, including SMAD4 alteration in LDN-193189 response and KRAS status in cetuximab resistance, but did not fully resolve discordant samples such as KRAS-mutant cetuximab-sensitive C45. Organoid-line-specific regulatory-network analysis identified heterogeneous post-treatment-associated rewiring. LDN-193189 yielded a 15-edge common sensitivity network that retained 12 edges at a more stringent ΔECv threshold, whereas cetuximab yielded a smaller two-edge stringent core when all organoids meeting the predefined high-sensitivity threshold were included. Full-cohort support analysis showed heterogeneous network support among moderate-response and resistant organoids, indicating that the common subnetworks should not be interpreted as validated monotonic classifiers across all response categories. qPCR follow-up provided node-level support for selected network-derived genes. Organoid-line-specific regulatory-network analysis identified interpretable candidate targeted-drug response-associated subnetworks in patient-derived CRC organoids beyond conventional genomic and expression-only comparators. These findings support an exploratory translational network-medicine framework for studying therapeutic heterogeneity in patient-derived CRC organoid drug-response studies.