The Switch That Took Forty Years to Flip

Every cell in the human body relies on molecular signal transduction pathways that dictate whether to proliferate, remain quiescent, or undergo apoptosis. Central to these networks is the RAS superfamily of small guanosine triphosphatases (GTPases)—proteins that act as binary molecular switches.

In normal cellular physiology, RAS cycles between two distinct states: an active state bound to guanosine triphosphate (GTP) and an inactive state bound to guanosine diphosphate (GDP). When a growth factor binds a cell-surface receptor, guanine nucleotide exchange factors (GEFs, such as SOS1) stimulate RAS to release GDP and bind intracellular GTP. In this active conformation, flexible regions called the switch I and switch II loops undergo a conformational change, forming a docking surface that recruits downstream effector kinases—chiefly RAF in the MAPK/ERK pathway and PI3K. Signaling is terminated when GTPase-activating proteins (GAPs, such as NF1) accelerate RAS's intrinsic ability to hydrolyze GTP back to GDP, returning the switch to its resting state.

When mutated, however, a single amino acid substitution can prevent GTP hydrolysis, locking the protein in the active GTP-bound state. Among human oncogenes, the KRAS isoform is the most frequently mutated, driving a substantial fraction of pancreatic ductal adenocarcinomas, colorectal cancers, and non-small cell lung carcinomas.

For four decades, KRAS was widely designated by cancer biologists as "undruggable." While crystallographers had resolved its three-dimensional structure at atomic resolution, the protein lacked deep, hydrophobic binding pockets suitable for conventional small-molecule inhibitors. Furthermore, its affinity for GTP is measured in the picomolar range—meaning it binds its cellular fuel roughly a thousand times more tightly than typical drug candidates could compete against. The hurdle was fundamentally biophysical rather than conceptual.

A Brief Primer on RAS Biology

Before the drug story can be told, the molecular vocabulary must be established. The following terms appear throughout.

RAS is a family of small GTPases — proteins that bind and hydrolyze guanosine triphosphate (GTP). In its active, GTP-bound state, RAS recruits downstream effector proteins such as RAF, MEK, and ERK, initiating a signaling cascade that drives cell proliferation and survival. In its inactive, GDP-bound state, RAS releases these effectors and the signal stops.

KRAS is the most frequently mutated member of the RAS family in human cancer. The most common mutations occur at three positions: G12 (glycine at position 12), G13, and Q61. Each mutation impairs GTP hydrolysis, leaving KRAS permanently GTP-bound and constitutively active.

G12C is a specific substitution in which glycine 12 is replaced by cysteine. Cysteine contains a sulfur atom, and sulfur atoms can form strong, irreversible bonds with certain chemical groups. This biochemical quirk would prove critical.

G12D, G12V, and other G12 substitutions are found in different tumor types. G12D is the dominant mutation in pancreatic cancer, present in over 90% of cases. G12C is most common in lung adenocarcinoma, comprising 25–32% of cases.

Farnesyltransferase is the enzyme that attaches a lipid anchor to RAS, allowing it to embed in the cell membrane. Early attempts to block this enzyme failed because the cell simply used a backup enzyme, geranylgeranyltransferase, to accomplish the same task.

The switch-II pocket is a shallow, transient indentation on the surface of GDP-bound KRAS G12C that appears only when the protein is in its inactive state. It sits adjacent to the mutant cysteine. This pocket was invisible in earlier crystal structures and would not be discovered until 2013.

The Geometric Problem

To understand why most successful cancer drugs work, it helps to know how they bind their targets. The vast majority latch onto a pocket or groove on a protein's surface — a physical indentation where a small molecule can nestle and block the protein's function.

RAS, however, binds its GTP signaling molecule with picomolar affinity. The connection is tighter than almost any drug could hope to compete with. Worse, its surface is unusually smooth. No deep pocket, no obvious groove, no handhold.

Early attempts to block RAS indirectly, such as farnesyltransferase inhibitors in the 1990s, failed because the cell is not a simple machine with a single on/off button. When researchers blocked the enzyme that helps RAS anchor to the cell membrane, the cell simply routed around the blockade using a backup enzyme. By the early 2010s, RAS had a well-earned reputation: fascinating to study, hopeless to drug. It became the textbook example of an "undruggable" target.

The Turn: A Hidden Pocket and a Covalent Bond

The turn came in 2013. Kevan Shokat's group at the University of California, San Francisco found something that had not shown up in any previous crystal structure of RAS: a shallow pocket that only appears when the mutant G12C protein is in its "off," GDP-bound state, right next to the mutant cysteine.

A cysteine is an amino acid that contains a sulfur atom, and sulfur atoms can form strong, irreversible bonds with certain chemical groups. Shokat's team realized they could use a small covalent fragment — a molecule designed to form a permanent chemical bond — to reach into that hidden pocket and weld itself there, trapping the switch in the off position.

It was a proof of concept, not a drug. The molecule was too weak and too unselective to give to a patient. But it reopened a field that had been written off, proving that RAS did have a vulnerability after all, and that the right chemistry could exploit it.

From Bench to Bedside: Sotorasib and Adagrasib

Eight years later, that discovery became a treatment.

Sotorasib, the first direct KRAS inhibitor, was approved in May 2021 for previously treated KRAS G12C-mutant lung cancer. Its approval was based on the CodeBreaK 100 trial, a Phase 1/2 study that would have been unthinkable just years earlier: a 37% response rate and 12.5-month median overall survival in a heavily pretreated population, published in the New England Journal of Medicine.

In oncology, "response rate" means the percentage of patients whose tumors shrank by a predefined amount. "Median overall survival" means half the patients lived longer than that figure and half lived shorter. For patients who had exhausted standard therapies, these numbers represented a genuine new option where none had existed before.

Adagrasib followed in December 2022, with a 43% response rate in the KRYSTAL-1 trial and a similar overall profile.

Both drugs belong to a class called covalent inhibitors, meaning they form an irreversible bond with their target. But adagrasib's label carries a heavier warning list than sotorasib's: liver enzyme elevations in roughly a third of patients, QT-interval prolongation (a potentially dangerous heart rhythm disturbance), and gastrointestinal toxicity. These side effects are a reminder that even a precisely targeted drug still has to move through a body that was not built with G12C-selectivity in mind. A drug can be exquisitely designed for one protein, but the liver, the heart, and the gut do not know that. They process the molecule as a foreign chemical, and sometimes they react accordingly.

The Confirmatory Trials: Modest Gains and a Harder Truth

The confirmatory Phase 3 trials told a more complicated story than the early single-arm studies had suggested.

In CodeBreaK 200, sotorasib produced a median progression-free survival of 5.6 months versus 4.5 months for docetaxel in second-line KRAS G12C NSCLC. The improvement in progression-free survival was statistically significant, but no significant overall survival difference was observed. Diarrhea at grade 3 occurred in 12% of patients; 27% required dose holds, and 11% discontinued.

In KRYSTAL-12, adagrasib achieved a median progression-free survival of 5.5 months versus 3.8 months for docetaxel. A notable feature was a 24% intracranial response rate. Nausea was the dominant patient-reported complaint — seen in 34% of patients, though mostly grade 1 to 2.

Both drugs target the G12C mutation in its inactive GDP-bound ("off") state. And both face a common resistance challenge: on-target mutations at the G12C binding pocket, off-target upregulation of wild-type RAS and other RAS family members, epithelial-to-mesenchymal transition, and histologic transformation. The same mechanisms that create resistance to one drug create resistance to the other, limiting sequential use.

The Resistance Problem: Two Escape Routes

Neither drug lasted as long as anyone hoped. Resistance set in within months in most patients, and later sequencing of resistant tumors turned up two broad patterns that illustrate a fundamental challenge in cancer biology.

Some cancers acquired second-site mutations — changes at residues like Y96 and H95, right in or near the drug's own binding pocket — that simply blocked the inhibitor from docking a second time. It is as if the tumor changed the lock after the first key was made.

Others did not touch KRAS at all. They reactivated the same growth signal through related proteins such as NRAS, BRAF, or amplified MET, routing around the roadblock rather than removing it.

Cancer cells are under intense evolutionary pressure to survive, and they are remarkably good at finding an alternative key, or an alternative door altogether.

Plasma biomarker analysis of CodeBreaK 100 revealed that acquired genomic alterations at disease progression were heterogeneous and polyclonal, including changes in multiple genes and RTK pathway dysregulation. Similarly, adagrasib treatment resulted in several mutations in the drug binding pocket. Notably, although H95 alteration showed reduced sensitivity to adagrasib, this did not affect sensitivity to sotorasib — a small window for sequential therapy that is rarely clinically useful.

The Next Generation: Better G12C Inhibitors

The current generation is trying to close more doors at once. Several next-generation KRAS G12C inhibitors with improved potency over the first-in-class agents are now in clinical trials.

Divarasib (GDC-6036), a highly selective covalent KRAS G12C inhibitor, demonstrated an ORR of 53.4% and a median PFS of 13.1 months in previously treated metastatic NSCLC — numerically higher than sotorasib or adagrasib, with grade ≥3 treatment-related adverse events in only 11% of patients. In the Krascendo-170 study, divarasib plus pembrolizumab achieved a 73% ORR and 19.3-month median PFS in the PD-L1-positive population. However, grade 3–4 toxicities occurred in 65.4% of patients, and 12.8% discontinued treatment.

Olomorasib (LY3537982) showed an ORR of 41% with an 8.1-month median PFS in patients previously treated with a KRAS G12C inhibitor. In combination with pembrolizumab, the LOXO-RAS-20001 trial demonstrated a 57% ORR, including 73.9% in treatment-naïve patients. The FDA granted olomorasib Breakthrough Therapy designation in September 2025 for first-line KRAS G12C NSCLC in combination with pembrolizumab, and a second designation in August 2026 for previously treated pancreatic cancer.

Elisrasib (D3S-001), an off-state inhibitor designed for rapid target engagement, achieved a 58% ORR and 12.2-month median PFS in previously treated KRAS G12C inhibitor-naïve NSCLC, and a remarkable 78% ORR as monotherapy in the treatment-naïve setting. In combination with pembrolizumab, the ORR reached 81.3%, with 95% in patients with PD-L1 expression ≥50%. One fatal case of myocarditis-ILD was reported.

MK-1084 achieved a 38% ORR as monotherapy and, in combination with pembrolizumab, a 77% ORR in PD-L1-positive patients with a median PFS of 25 months.

The Deeper Shift: RAS(ON) Inhibitors

The most consequential development in the KRAS field is not a better G12C inhibitor. It is a fundamentally different strategy.

Daraxonrasib (RMC-6236), a RAS(ON) inhibitor, represents a new mechanistic class. Instead of locking onto one inactive mutant in its "off" state, it binds active, GTP-loaded RAS across several mutant forms. Rather than waiting for the switch to flip off and then jamming it, it grabs the switch even while it is flipped on.

The mechanism is elegant. Daraxonrasib forms an inhibitory tri-complex with cyclophilin A (CyPA) and active RAS-GTP. Cyclophilin A is a ubiquitous cellular protein that normally has nothing to do with RAS signaling. The drug acts as a molecular glue, bringing CyPA into contact with active RAS and physically blocking effector recruitment. Because the drug binds to a surface that is present in all active RAS proteins — not just G12C — it can inhibit multiple RAS variants simultaneously.

In 2026, daraxonrasib reported results that would have been unthinkable a decade ago. In the Phase 3 RASolute-302 trial, it significantly extended overall survival compared with chemotherapy in previously treated metastatic pancreatic cancer. Pancreatic cancer is a disease where KRAS mutations, mostly G12D rather than G12C, are found in over 90% of tumors. It is also one of the deadliest cancers, with a five-year survival rate that has historically hovered in the single digits, and where no KRAS-directed therapy had worked at all until now.

The trial enrolled 500 patients with previously treated metastatic pancreatic ductal adenocarcinoma, randomized to daraxonrasib or investigator's choice of chemotherapy. The results, presented at the 2026 ASCO Annual Meeting and published in the New England Journal of Medicine, were striking:

The 12-month overall survival rate was 53.3% with daraxonrasib versus 18.7% with chemotherapy. These results were consistent across the intent-to-treat population, which included patients with various RAS mutations and even those without an identified tumor RAS mutation.

Daraxonrasib also demonstrated a statistically significant delay in time to deterioration in cancer-related pain and quality of life compared with chemotherapy. The FDA authorized an expanded access program in May 2026, and Revolution Medicines intends to submit a New Drug Application under a Commissioner's National Priority Voucher.

Beyond G12C: G12D and the Degrader Frontier

The expansion beyond G12C is the defining trend of the current KRAS landscape. G12C inhibitors opened the market, but they address only a portion of KRAS-mutated cancers.

G12D is the most common KRAS mutation in pancreatic cancer and the second most common in colorectal cancer. Unlike G12C, G12D does not introduce a reactive cysteine that can form a covalent bond. This makes it inaccessible to the same covalent chemistry that cracked G12C. Two approaches are now showing proof-of-concept.

Setidgrasib (ASP3082), a KRAS G12D-directed protein degrader from Astellas Pharma, achieved a 22% ORR in early clinical data. In April 2026, Astellas dosed the first patient in a Phase 3 registrational study evaluating setidgrasib in combination with chemotherapy for first-line treatment of KRAS G12D-mutated metastatic pancreatic cancer. A protein degrader does not merely inhibit KRAS — it recruits the cell's own disposal machinery to destroy it entirely.

A separate KRAS G12D protein degrader (not yet named in public disclosures) achieved a 36% ORR in a study published in the New England Journal of Medicine in 2025 — the first published clinical data for a G12D-targeted therapy.

RMC-6291 (elironrasib), a G12C-selective RAS(ON) inhibitor from Revolution Medicines that uses the same cyclophilin A tri-complex mechanism as daraxonrasib, is in clinical trials. Unlike daraxonrasib, which is pan-RAS, RMC-6291 is selective for G12C — offering a more targeted approach for tumors where G12C is the sole driver.

The Graveyard Is a Map

The KRAS story is less a single triumph than a slow demonstration that "undruggable" was a statement about the tools available in a given decade, not about the target itself. The switch did not change. The protein's structure has been the same for millions of years of evolution. What changed was our ability to see the fleeting shape it takes on the way to flipping off — and, more recently, to hold on to it even while it is flipped on.

The shift is clear: from blocking the enzyme that processes RAS (farnesyltransferase), to blocking the pathway downstream of RAS (MEK), to covalently trapping one specific mutant in its inactive state (G12C OFF), to non-covalently blocking multiple active mutants simultaneously (pan-RAS ON), to destroying the mutant protein entirely (degraders). Each step represents a deeper understanding of what the biology actually requires.

What Comes Next

The KRAS story is still unfolding, but several patterns are already clear.

First, "undruggable" was always about chemistry, not biology. KRAS was never invincible. It was simply incompatible with the drug design paradigms of the 1990s and 2000s. The discovery of the switch-II pocket, the development of covalent fragment screening, and the invention of molecular glue mechanisms each opened doors that previous approaches could not.

Second, resistance is the central challenge, and it is not solved by better G12C inhibitors alone. The next-generation OFF-state inhibitors produce higher response rates and longer PFS, but they face the same evolutionary escape routes. The real solutions will likely come from combinations — RAS inhibitors plus SHP2 inhibitors, plus immunotherapy, plus downstream pathway blockade — and from mechanisms that bypass resistance entirely, such as pan-RAS inhibition and protein degradation.

Third, the competitive landscape is fragmenting by mutation and mechanism. G12C is now a crowded field with multiple approved and near-approved agents. G12D is the next frontier, with protein degraders leading and direct small-molecule inhibitors still struggling with the lack of a reactive cysteine. Pan-RAS ON-state inhibition is the broadest approach but requires validation beyond pancreatic cancer.

Fourth, the biggest unmet need remains pancreatic cancer. With over 90% of cases driven by KRAS mutations — mostly G12D — and a five-year survival rate in the single digits, an effective KRAS-directed therapy would transform the disease. The RASolute-302 results suggest this may finally be within reach.

The switch did not change. What changed was our ability to see it, to bind it, and to hold it — first in one fleeting conformation, then in another, and now in the very act of signaling. Somewhere in that same protein surface where KRAS has been doing its ancient work, the next round of therapies is being tested against it, more precisely aimed than anything that came before.