The regulatory wins seemed almost anticlimactic. By the time the FDA cleared sotorasib paired with panitumumab this January—the second KRAS G12C combination to reach patients with colorectal cancer in seven months—the oncology world had already shifted its gaze to tougher puzzles. G12C, after all, accounts for barely 13% of KRAS mutations in lung cancer and even less in pancreatic and colorectal tumors. The real prize, the mutation that has haunted drug developers for decades, is G12D.
It was supposed to be impossible. Until Revolution Medicines made it look almost routine.
On January 8, the California biotech secured Breakthrough Therapy Designation for zoldonrasib in KRAS G12D-mutated non-small cell lung cancer, backed by a 61% objective response rate at the 1200 mg dose. The company's RAS(ON) inhibitor platform—which recruits a cellular chaperone protein called cyclophilin A to trap mutant KRAS in an inactive state—represents the kind of allosteric wizardry that five years ago was confined to academic whiteboards. Now Revolution is dosing patients with pancreatic ductal adenocarcinoma and assembling combination trials that pair zoldonrasib with checkpoint inhibitors and the company's own multi-selective KRAS inhibitor, RMC-6236.
The KRAS breakthrough is more than a scientific milestone. It's emblematic of a broader shift rippling through cancer drug development, where targets once waved off as "undruggable" are surrendering to an arsenal of novel chemistries, protein degraders, conjugated payloads, and machine learning-accelerated design. Global oncology spending hit $252 billion in 2024, according to IQVIA, and is on track to reach $441 billion by 2029. Much of that growth is being captured by modalities that didn't exist—or barely existed—a decade ago: antibody-drug conjugates, bispecifics, radiopharmaceuticals.
The question now isn't whether the impossible can be done. It's how fast it can be scaled, and whether the business model can keep up with the science.
Why G12D Matters (And Why It Almost Didn't Happen)
Kevan Shokat, the UCSF chemist whose lab cracked KRAS G12C's code, has described G12D as "dominant and harder to target." The mutant protein adopts conformations that shrug off the covalent warhead strategies that worked so elegantly against G12C. Revolution's solution was to abandon that playbook entirely, instead targeting the active, GTP-loaded state and using cyclophilin A as molecular glue to lock the complex in place.
Bristol Myers Squibb, which acquired Mirati Therapeutics and its MRTX1133 G12D program, pulled the plug on that Phase 1/2 trial shortly after the acquisition closed, citing formulation and pharmacokinetic headaches. The setback didn't invalidate the target—it underscored just how chemically finicky these molecules can be. Revolution now holds what appears to be the clearest shot at a G12D approval, possibly by the end of the decade if pivotal data cooperates.
Meanwhile, companies like Erasca and Boehringer Ingelheim are going after the problem upstream, testing inhibitors of SHP2 and SOS1 to block RAS activation before it starts or to counter acquired resistance to direct KRAS inhibitors. The combination paradigm—a RAS inhibitor stacked with upstream interference or immune checkpoint blockade—is quickly becoming standard operating procedure.
Transcription Factors: The Next Frontier (Or Maybe Just the Next Headache)
If KRAS was tough, transcription factors were thought to be outright impossible. They lack the deep hydrophobic pockets that let traditional small molecules latch on. No pocket, no drug. Except now three different transcription factors are showing clinical signals, and at least one has already crossed the finish line.
p53, mutated in more than half of all cancers, became targetable when researchers identified a specific stabilizing mutation—Y220C—that creates a druggable pocket. PMV Pharmaceuticals' rezatapopt is a small molecule designed to restore wild-type p53 function in tumors carrying that mutation. The company has pivotal studies underway, angling to become the first reactivation therapy for a dominant tumor suppressor. Whether it works at scale remains to be seen, but the fact that it exists at all would have seemed absurd a few years ago.
Menin, a transcriptional regulator implicated in certain aggressive leukemias, became the second validated transcription factor target when Syndax's revumenib (Revuforj) won FDA approval last November for relapsed or refractory acute leukemia with KMT2A translocation. The oral menin inhibitor disrupts the menin-MLL fusion complex, coaxing blasts back toward differentiation.
TEAD transcription factors—central players in the Hippo pathway and implicated in mesothelioma and other malignancies—represent the next test case. Vivace Therapeutics' VT3989, a TEAD palmitoylation inhibitor, is showing early activity in mesothelioma and has secured both orphan drug and fast-track designations. The company plans to kick off a Phase 3 trial this year. Novartis, for its part, discontinued its competing TEAD program, IAG933, late last year. A reminder, perhaps, that mechanism validation remains fragile terrain.
Then there's STAT3, a master regulator of oncogenic signaling and immune suppression. Tvardi's oral inhibitor TTI-101 has generated early responses across multiple tumor types, and academic labs are now engineering STAT3 degraders using PROTAC and D-PROTAC technologies. Will it translate to approvals? Too early to say. But the fact that serious money and serious science are chasing it says something.
The Platform Play: Degradation, Glues, and Everything in Between
The leap from "undruggable" to "difficult-to-drug" rests on three converging platform technologies: targeted protein degradation, molecular glues, and next-generation conjugates.
Targeted protein degraders—PROTACs, in the jargon—hijack the ubiquitin-proteasome system to eliminate disease proteins entirely rather than merely inhibiting them. None have won FDA approval yet, but Arvinas and Pfizer's vepdegestrant came tantalizingly close. The oral estrogen receptor degrader met its primary endpoint in the ESR1-mutant cohort of the VERITAC-2 Phase 3 trial, showing progression-free survival benefit in ER-positive, HER2-negative breast cancer. The intention-to-treat population result was messier, and the path to approval will hinge on overall survival data and some careful regulatory maneuvering. Still, it's the first Phase 3 win for a degrader—proof, at least, that the modality can deliver clinical differentiation.
Molecular glues, conceptually simpler cousins of PROTACs, are also generating early proof-of-concept. Monte Rosa Therapeutics' MRT-2359, a GSPT1 degrader, showed PSA declines in metastatic castration-resistant prostate cancer when combined with enzalutamide, targeting MYC-driven tumors. The company's VAV1 glue, MRT-6160, attracted a deal with Novartis. SEED Therapeutics dosed its first patient with an RBM39 glue, ST-01156, just last month—barely weeks after clearing IND—backed by partnerships with Eli Lilly and Eisai.
The glue platform is pushing beyond the well-trodden cereblon and VHL E3 ligases. Recent academic reviews in Cancers and Nano Today highlight emerging autophagy-mediated degradation strategies—LYTACs, AUTACs, ATTECs—that could unlock entirely new classes of both intracellular and extracellular targets. Whether that happens in five years or fifteen is anyone's guess.
When Machines Start Designing Molecules
Computational design is no longer a side project. McKinsey estimates generative AI could deliver $4 billion to $7 billion in annual operational value across biopharma R&D, with pharma's biggest players moving from cautious pilots to enterprise-scale deployment. A January 2026 Nature Reviews Cancer survey documented AI agent use across oncology workflows, from target identification to trial design optimization.
Insilico Medicine put an AI-designed pan-TEAD inhibitor into the clinic, demonstrating end-to-end computational drug discovery. The molecule was generated by neural networks trained on protein structure and chemical space, then optimized for ADME properties and selectivity—all before anyone synthesized it. Whether it succeeds clinically is, in some ways, beside the point. The proof-of-concept is what matters: machine learning can now propose chemically novel scaffolds for historically intractable targets.
The FDA's Project Optimus initiative, which mandates dose optimization in oncology trials, and the agency's 2025 draft guidance on radiopharmaceutical dosage optimization both signal that regulators expect sponsors to use every available tool—AI-driven modeling included—to bring forward optimized regimens from the start. No more throwing three doses at the wall and seeing what sticks.
The Money Follows the Science (Mostly)
Antibody-drug conjugates are the most commercially validated of the novel modalities. Multiple market forecasts project robust double-digit growth through the 2030s, driven by HER2-low expansion (Enhertu), TROP2 targeting (Dato-DXd, approved in January for HR-positive, HER2-negative breast cancer), and next-generation payloads. Bicycle Therapeutics' BT8009, a Nectin-4-targeted bicycle drug conjugate using a novel BDC scaffold, has fast-track designation in urothelial cancer and could differentiate from MMAE-based ADCs through improved tumor penetration and bystander effects.
Radiopharmaceuticals are seeing similar momentum. Novartis expanded Pluvicto (Lu-177-PSMA) to earlier lines of metastatic castration-resistant prostate cancer last March, and interim data from metastatic hormone-sensitive prostate cancer trials suggest the modality may migrate even earlier in the disease course. The radioligand therapy market is projected to hit $13.5 billion by 2032. Major M&A—Bristol Myers Squibb acquiring RayzeBio, Eli Lilly buying Point Biopharma, AstraZeneca snapping up Fusion—validates the sector's strategic importance.
But policy headwinds are complicating the picture. The Inflation Reduction Act's small-molecule penalty—a shorter exclusivity window before Medicare price negotiation kicks in, compared to biologics—is reshaping portfolio strategies in ways both subtle and stark. Academic analyses published in Health Affairs Scholar and discussed at ASCO last year show a 45.3% decline in post-approval small-molecule oncology trials since the IRA's passage, versus flat or growing biologic trial starts. The distortion favors large molecules, ADCs, and earlier combination registrations. Which may or may not be optimal from a patient care perspective.
BreakThru Medicine, which emerged from stealth last month with a $60 million Series A, embodies the sector's current thesis. The company's founders—Steve Potts (OncoMyx, Ignyta), Mark Mulvihill (Warp Drive Bio, HiberCell), and Brian Barnett, a Genentech ADC launch veteran—are running multiple independent platforms targeting small molecules, ADC novel payloads, and molecular glues. Potts, who testified before Congress in 2023 about the IRA's impact, has been vocal about the distortions. The company aims to put its first small molecule into the clinic in early 2027, emphasizing what Potts calls a tumor-agnostic, high-bar approach to breakthrough potential. Whether the economics will support that strategy over the long haul is an open question.
Execution Still Matters More Than Chemistry
Clinical execution remains the bottleneck, and perhaps always will be. Vivace's VT3989 is advancing while Novartis abandoned its TEAD program. Revolution's zoldonrasib is thriving while Mirati's MRTX1133 flamed out post-acquisition. PMV's p53 reactivator is in pivotal studies, but most transcription factor-targeted programs are still in early phases, where attrition is brutal and unforgiving.
The multi-modal strategies companies are pursuing—molecular glues for transcription factors, covalent inhibitors for kinases, degraders for nuclear receptors, conjugates for solid tumors—each carry modality-specific risks that can torpedo even the most elegant chemistry. Degraders must achieve sufficient ubiquitination without off-target depletion. Glues require exquisite ternary complex stability. ADCs depend on linker chemistry, payload potency, and tumor penetration. Radiopharmaceuticals face manufacturing and logistics complexity that makes traditional small molecules look simple by comparison.
And none of that accounts for the messy reality of patient biology: tumor heterogeneity, acquired resistance, pharmacokinetic variability. The science can be perfect. The execution can still fail.
What Happens Next
The next 24 to 36 months will likely deliver the first KRAS G12D approval, the first PROTAC approval (pending vepdegestrant's overall survival data and some regulatory maneuvering), and potentially the first TEAD inhibitor approval if Vivace's Phase 3 succeeds. Tumor-agnostic approvals, which reached eight therapies across nine indications by 2025, will expand as molecular diagnostics become routine and basket trials prove their value beyond niche populations.
The term "undruggable" is fading from the vocabulary, replaced by "difficult-to-drug" or, more optimistically, "not yet drugged." KRAS G12D, once the poster child for chemical impossibility, now has a Breakthrough Therapy-designated molecule in pivotal development. p53, the "guardian of the genome" and the most frequently mutated gene in cancer, has a reactivation therapy in late-stage trials. TEAD, buried deep in the nucleus without an obvious binding site, is yielding to palmitoylation inhibitors that most chemists wouldn't have even thought to try a decade ago.
The convergence of platform innovation, computational design, and relentless chemical ingenuity is rewriting what's possible. What was impossible ten years ago is difficult today. What's difficult today may be routine tomorrow.
Assuming, of course, that clinical execution, regulatory alignment, and capital patience all hold. Which is never a given.
