Oncogenic RAS as an adaptive signaling ecosystem: context-dependent vulnerabilities and resistance-informed therapeutic design

Background

Direct RAS targeting has progressed from pharmacologic proof of concept to routine treatment for selected KRAS G12C-mutant cancers. KRAS G12C inhibitors and KRAS G12C-EGFR combinations established allele- and lineage-specific treatment paradigms. More recently, the multiselective RAS(ON) inhibitor daraxonrasib became the first FDA-approved direct RAS-targeted therapy for metastatic pancreatic adenocarcinoma, while the KRAS G12D-directed degrader setidegrasib remains investigational. Target engagement alone, however, does not ensure durable disease control, and the same drug class can perform differently across tumor lineages. Main body: This Review develops an adaptive RAS signaling ecosystem model in which allele and isoform biology, membrane organization, feedback architecture, mutant-wild-type RAS crosstalk, co-mutation patterns, chromatin state, epithelial-mesenchymal plasticity, metabolism, spatial heterogeneity, the tumor microenvironment and treatment delivery jointly determine pathway dependence and therapeutic escape. Comparisons among pancreatic ductal adenocarcinoma, colorectal cancer, lung adenocarcinoma and NRAS-mutant melanoma show how related oncogenic RAS nodes acquire different vulnerabilities in distinct tissue contexts. We integrate direct RAS inhibition with mechanism-guided vertical and horizontal strategies directed at feedback rebound, bypass signaling, cell-state persistence, stromal protection, metabolic adaptation and immune exclusion. We then translate this model into a biomarker-integrated development framework that links serial circulating tumor DNA, modality-matched pharmacodynamic assays, paired tissue sampling, spatial profiling and toxicity-aware scheduling to prespecified, falsifiable treatment hypotheses.