Allosteric and covalent strategies for RAS-driven cancers from KRAS G12C to pan-RAS inhibition

RAS was regarded for nearly three decades as the archetypal “undruggable” therapeutic target. Its picomolar affinity for nucleotides, the millimolar concentration of intracellular GTP, and the absence of an apparent ligandable pocket on its protein surface made direct pharmacological inhibition seem not merely challenging but virtually infeasible. This prevailing view was ultimately overturned by the identification of a cryptic pocket beneath the Switch II region and the recognition that a mutation-introduced cysteine could provide a tractable site for covalent inhibitor engagement. This review traces what has followed. Covalent OFF-state inhibitors validated mutant RAS as a drug target and reached the clinic, but they remain confined to KRAS G12C, an allele that is common in lung cancer and scarce in the pancreatic and colorectal tumours where RAS mutation is nearly universal. The field's response has been to stop waiting for the inactive state. Non-covalent allele-selective agents engage Switch II without a reactive handle; tri-complex molecular glues co-opt cyclophilin A to occlude the effector interface on active RAS across alleles and isoforms; targeted degraders remove the protein outright; nucleotide-free trappers intercept the apo intermediate during exchange. Randomised Phase III data now support pan-RAS inhibition in pretreated metastatic pancreatic cancer. Resistance, predictably, has followed, and its architecture is instructive: on-target pocket and glue-interface mutations, effector-level competition, copy-number gain, and non-genetic bypass through YAP/TAZ-TEAD and lineage plasticity. We synthesise structural, preclinical, and clinical evidence across these modalities, examine where the data remain thin, and argue that durability will depend less on potency than on biomarker-guided combinations chosen against the escape route a given tumour is likely to take.