on forty years of failure and one spring of success: the protein with no handle, the chemist who found one, the pancreatic-cancer trial that doubled survival, the five-week approval, the $477,600 price, the $30 billion talks that collapsed — and a clear-eyed map of who wins, who loses, and how to think about capitalizing on it without fooling yourself.* **By Lily Caruso · Saturday, September 5, 2026 — long weekend edition, fact-checked before publication. Not investment advice. Every number carries its source.** --- ## Prologue: a number that wouldn't move For three years running, the American Cancer Society has published the same five-year survival rate for pancreatic cancer: **13 percent.** Every other major cancer has crept upward — all cancers combined crossed 70 percent this year — and pancreatic cancer just sat there, "the deadliest major cancer and the only one with a five-year survival rate below 20 percent," on its way to becoming the second-leading cause of cancer death in America by 2030.
This year it will be diagnosed in about 67,530 people and will kill about 52,740 of them. On August 26, 2026, the FDA approved a pill that doubled median survival in patients whose first treatment had already failed. It did so in five weeks, under a program most investors had never heard of, for a company that had walked away from takeover talks at roughly $30 billion seven months earlier.
The drug attacks a protein that ninety-two percent of pancreatic tumors depend on — a protein the pharmaceutical industry spent four decades calling undruggable. I want to tell that story properly, science first, because the science is the reason the money exists. Then I want to be honest about the money, because a $45 billion company with one approved drug and a $644 million quarterly loss is a place where honesty gets expensive fast. --- ## Part I — The science ### The switch with no handle Every cell in your body carries a family of proteins called RAS — KRAS, NRAS, HRAS — that work as on-off switches for growth.
Bound to a molecule called GTP, the switch is on and the cell divides. Bound to GDP, it's off. Cancer's oldest trick is to break the switch in the on position, and it does that most often at a single spot, the twelfth amino acid, with mutations oncologists know by their coordinates: G12D, G12V, G12C, G12R.
The scale is hard to overstate. In a Foundation Medicine dataset of 426,706 patients, **23 percent of all adult cancers** carried a KRAS alteration. In pancreatic cancer it was 92 percent; in colorectal cancer, 49; in non-squamous lung cancer, 35.
KRAS isn't a cancer. It's the engine inside nearly a quarter of them. And for forty years nobody could touch it.
Kevan Shokat, the UCSF chemist who eventually did, describes the problem simply. The protein's surface was "relatively smooth, with only one notable binding pocket," and that pocket held its natural partner so tightly that "no drug could outcompete it." You can't block a door that's bolted from the inside. ### The cysteine trick, and its limit In 2013, Shokat's lab noticed something about one particular mutation. G12C swaps in a cysteine, and, in his words, "whenever chemists see a cysteine, regardless of what the protein looks like, there's a chemical opportunity to make an irreversible bond to the target." His team found a hidden pocket that exists only when KRAS is switched off, and built a molecule that wedges into it and welds itself to the cysteine.
Eight years later that idea became Amgen's sotorasib, approved in May 2021, and Mirati's adagrasib, approved in December 2022. Shokat: "I thought that would be the 10-year plan, but it's here already." Here is the limit, and it matters for everything that follows. The trick only works on G12C — and G12C is the *minority* mutation.
Across cancers, G12D accounts for 29 percent of KRAS mutations, G12V for 23, and G12C for just 15. In pancreatic cancer, G12D alone is 43 percent and G12C is rare. The first-generation drugs opened one door in seven.
They also work only when the switch is off, and a cancer cell's mutant KRAS spends most of its time on. That's the science behind a commercial disappointment. Sotorasib did $111 million last quarter and adagrasib $55 million, on response rates of 30 to 40 percent and roughly six extra months before the tumor learns its way around the drug. ### The glue The company that opened the other six doors is Revolution Medicines, and its idea is worth understanding, because it's the reason the pancreatic trial worked.
Instead of waiting for KRAS to switch off, its drug — daraxonrasib, now sold as Rasonque — grabs the protein while it's *on.* It does that by acting as a "molecular glue": it recruits an abundant bystander protein inside the cell and sticks it to active RAS, forming a three-part complex that physically stops RAS from talking to anything downstream. Because the glue doesn't need a cysteine, it works across G12D, G12V, G12C, G12R and others. The company calls the class RAS(ON) inhibitors.
I'd call it the first time anyone reached the switch in the position it's actually in. ### What the trial showed The Phase 3 trial is called RASolute 302. Five hundred patients with metastatic pancreatic cancer who had already progressed on first-line treatment were randomized to daraxonrasib — 300 milligrams once a day, by mouth — or to the chemotherapy their doctor would have chosen anyway. The data cutoff was February 10; Brian Wolpin of Dana-Farber presented the results at ASCO this spring.
In the RAS G12 population, the primary endpoint, **median overall survival was 13.2 months against 6.6 months** on chemotherapy — a hazard ratio of **0.40**, a 60 percent reduction in the risk of death, with a p-value of 5.9 × 10⁻¹⁰. At twelve months, 53 percent of the drug's patients were alive, versus 19 percent on chemo. Progression-free survival was 7.3 months versus 3.5, hazard ratio 0.45.
The response rate was 33 percent versus 12. Every one of those endpoints reached statistical significance at the *first* interim analysis, and the results across the full trial population — including the roughly forty patients without a G12 mutation — were nearly identical. Two things about the quality of that survival.
Patients on the drug stayed on it for a median of 6.2 months, versus 1.5 to 3.2 months on chemotherapy, and only 1.2 percent stopped for side effects, against 11.2 percent. And they hurt less: time to worsening pain was 9.2 months versus 3.8. Severe treatment-related side effects were actually *lower* on the drug — 43.6 percent versus 57.5 — though its own toxicities are real and specific.
Rash in 86 percent of patients, severe in 10 percent. Mouth sores in 57 percent. Dose reductions in 36 percent, mostly for the rash.
One patient died of treatment-related lung inflammation. Dr. Pashtoon Kasi of City of Hope gave Newsweek the sentence every oncologist I've read has some version of: *"This is not a cure, it's one more option for these patients.
But it's the best option we've ever had."* ### What it changes about treating cancer This is the part of the story that will outlast the stock, and I want to give it its due. *First, the label has no genetic test attached.* The FDA approved Rasonque for adults with metastatic pancreatic adenocarcinoma who have had at least one prior therapy, or who aren't candidates for multi-drug chemotherapy — full stop, no companion diagnostic. The benefit held across the whole trial population, and the label reflects that. In a cancer where 92 percent of patients carry the mutation, the test would have been a delay more than a safeguard.
It's a precedent: the first RAS drug is effectively biomarker-agnostic in its first indication. *Second, pancreatic-cancer care is about to change from the back of the line forward.* The approval is second-line. But the company already has a first-line trial (RASolute 303) and an adjuvant trial (RASolute 304, for patients whose tumors were surgically removed) under way, and its G12D-specific sibling, zoldonrasib, is in first-line combinations with chemotherapy and in a doublet with daraxonrasib itself. Fierce Pharma adds a wrinkle worth knowing: roughly a quarter of metastatic pancreatic patients never receive chemotherapy at all, and the "not a candidate for multiagent therapy" language in the label reaches them today.
The line between second-line and first-line is already blurrier than the headline. *Third, lung cancer is the real prize, and it reads out next year.* The American Cancer Society expects about 229,410 new lung cancers this year, roughly three-quarters of them non-small cell, and about a third of the non-squamous ones carry KRAS. The company's lung trial, RASolve 301, is expected to finish enrolling this year with an initial readout in 2027, and it already holds Breakthrough Therapy designation. Zoldonrasib's Phase 1 in previously treated G12D lung cancer showed a 52 percent confirmed response rate, a 93 percent disease-control rate and 11.1 months of progression-free survival, with no grade 4 or 5 events — numbers that would have been unthinkable for a G12D drug three years ago. *Fourth, the whole industry is now attacking the switch from every side.* Roche's divarasib just beat the incumbent G12C drugs in a 338-patient head-to-head Phase 3, on progression-free survival *and* overall survival.
Lilly's olomorasib and Merck's calderasib are in first-line lung Phase 3s alongside Keytruda. In April, Astellas put the first KRAS *degrader* — a molecule that destroys the protein rather than blocking it — into a 600-patient first-line pancreatic Phase 3. Erasca has a pan-RAS glue in Phase 1 with early responses at 8 milligrams.
And the field learned something from a failure: Elicio's KRAS vaccine missed its primary endpoint in June, which tells you round one went to the small molecules and round two will be about combinations. *Fifth — the honest caveat — resistance is coming.* Every targeted cancer drug eventually meets a tumor that evolves around it; the first-generation G12C drugs met theirs within months. The RAS(ON) drugs have the advantage of hitting the active state and of being combinable with one another and with chemotherapy. But a median survival of thirteen months is a floor to build on, not a ceiling to celebrate.
The next decade of this story is about what gets stacked on top of the glue. ### The five-week approval One more note, because it will echo through every biotech valuation for years. Revolution's application was accepted on July 22 and approved on August 26 — thirty-five days — under the FDA Commissioner's National Priority Voucher pilot, a program that compresses a ten-to-twelve-month review into weeks for medicines the agency deems national priorities. Acting Commissioner Kyle Diamantas: "It is our fundamental duty to deliver more cures and meaningful treatments to patients as quickly as possible." Whatever you think of the politics of the current FDA, a five-week oncology approval on full — not accelerated — data is now a precedent every board in the industry will cite. --- ## Part II — The business ### The company that walked away from Merck On January 8, the Financial Times reported that Merck was in talks to buy Revolution Medicines for $28 to $32 billion.
On January 26 the talks ended. The companies, per the Wall Street Journal, "were unable to agree on valuation." Merck needed the deal — it faces roughly $18 billion of patent-cliff erosion over five years — and Revolution's management, in effect, bet that the trial data would be worth more than the bird in hand. They were right.
The topline came on April 13. The stock, which had traded as low as $38.49 in the past year, closed Friday at **$210.02** — a market capitalization of **$45 billion**. As of late August it was up 178 percent on the year, against an S&P 500 up less than 13.
Twenty-two analysts rate it a Strong Buy, with an average target of $248.77 and a high of $320. Now the harder numbers. Revolution had **$3.9 billion** in cash on June 30, after raising $1.725 billion in stock and $500 million in convertible notes in April, plus $250 million from Royalty Pharma with up to $1.5 billion more available on milestones.
Its second-quarter net loss was **$644 million**, its R&D bill $395 million, and its full-year operating-expense guidance is **$2.1 to $2.2 billion.** This is a company spending like a large pharma with the revenue of a startup — on purpose, because the window to own RAS is now. ### What the drug is worth The Street's builds disagree in an instructive way. **RBC Capital Markets** projects **$11.5 billion in peak sales**, with $28 million in the current quarter ramping to $1.1 billion by the end of 2027. **Evercore ISI** models about **$2.4 billion** of Rasonque sales next year and roughly **$15.1 billion** from pancreatic cancer alone by 2034. An independent market model I read, from Inflection Labs, gets **$5.3 billion** in its base case — and it assumed a $200,000 annual price. The actual list price is **$39,800 per thirty days, about $477,600 a year**, more than double that assumption, with roughly 84 percent realized after discounts.
Adjust for the real price and you can see how RBC gets to eleven figures. Now do the arithmetic the market is doing. At $45 billion, the company trades at about four times RBC's peak sales — for a drug with one indication, launched by a company that has never sold anything, into a disease where 60 percent of patients are treated in community practices that have never managed this drug's rash.
An Insider Monkey analysis put it bluntly: "at $47 billion, the market has priced a flawless solo launch into oncology's toughest cancer." Insiders sold about $21.5 million of stock in the last ninety days and bought none. Short interest is about 6 percent of the float. Institutional ownership sat at 105 funds, flat on the quarter.
None of that is disqualifying. All of it is the crowd's imagination, priced. This desk has printed the statute all week — **the tape grades the guide against the crowd's imagination, not the estimate** — and it applies here with unusual force.
Rasonque's first launch numbers should arrive with the third-quarter report in early November. They won't be graded against RBC's $28 million. They'll be graded against what a $45 billion valuation implies, and those are very different bars. ### The exchange rate Two days before I wrote this, Ultragenyx lost 46 percent of its value in a morning because a Phase 3 in Angelman syndrome came back negative.
Revolution has gained five-fold this year because a Phase 3 in pancreatic cancer came back with a hazard ratio of 0.40. Same industry, same week, same rule, opposite signs. **Biotech's only currency is data**, and this is what the exchange rate looks like in both directions. Anyone who wants the upside of the second has to be able to survive the first. ### Who else wins — and who loses *The incumbents are the losers.* Amgen's Lumakras and Bristol's Krazati just lost a head-to-head Phase 3 to Roche's divarasib on both progression and survival; the RAS(ON) drugs are coming for their G12C patients from the other direction; and Citeline's Neha Anand had already capped the whole G12C category at "$3 billion to $4 billion by 2028" before either of those things happened.
For Amgen and Bristol these are rounding errors on the income statement. For the thesis, they're the tell: the first generation is being replaced before it ever got big. *Roche* has the best next-generation G12C drug and, through Foundation Medicine, one of the largest testing franchises that every G12C drug depends on. *Lilly* has olomorasib in two first-line lung Phase 3s and just received Breakthrough status in G12C pancreatic cancer. *Merck* has calderasib in a 675-patient first-line trial with subcutaneous Keytruda — and the distinction of being the buyer that walked away. *Astellas* owns the degrader bet. *BeOne* just took rights to all four of Revolution's RAS(ON) drugs across much of Asia (Japan and Korea excluded) and will fund a global Phase 3 itself — a capital-light way for a Chinese-rooted oncology company to own the next decade in Asia's biggest markets. *The small caps are optionality, priced accordingly.* Erasca's pan-RAS glue has early data and readouts due in the second half of this year and next. Elicio, whose vaccine missed, has cash into the fourth quarter and needs a partner.
Jacobio and the other Chinese G12C makers have domestic approvals and domestic prices. --- ## Part III — Where the money goes, and how to think about it honestly TrendyVest doesn't tell you what to buy. What we do is decompose the exposure, put odds on the things that resolve it, and name the evidence that would make us wrong. Here is the RAS trade, taken apart. **Tier one — the pure play, and what you're actually buying.** Owning Revolution at $45 billion means buying three things: a pancreatic launch that has to go nearly perfectly; a lung readout in 2027 that decides whether this is a $5 billion drug or a $15 billion franchise; and an option on a takeover that is worth *less* than it was in January, because the price now sits above what the most motivated buyer would pay.
The falsifiers are specific and dated — the November quarter (does the 2,000-patient expanded-access base convert to paid prescriptions?), first-line label progress, community-oncology adoption, and the RASolve 301 lung data. If you own this, you own a calendar. **Tier two — the large-cap way in.** Roche, Lilly, Merck and Astellas each carry a RAS program inside a diversified income statement. You give up the five-fold upside and you buy the thing Revolution doesn't have: the ability to lose a trial and survive it.
The head-to-head result already in hand favors Roche. The first-line lung race is a three-way that resolves in 2027 and 2028. **Tier three — the diversified instrument.** The equal-weight biotech ETF, XBI, closed Friday at $163.81, up 75.7 percent over the past year and near the top of its $92.49-to-$169.89 range. Two honest notes.
Biotech is a duration asset — the sector that trades most like a long bond — and this desk spent the week documenting a 2-year yield at a 52-week high and a 58 percent probability of a rate hike on the 16th. The sector's macro falsifier is next Friday's CPI, not any trial. And a 75 percent year means the easy money in the basket has already been paid. **Tier four — the hedge, properly understood.** I was asked where to hedge.
The honest hedge in biotech isn't an instrument; it's a discipline. Size a position for the outcome that would cut it in half, because in this sector that's a Thursday, not a hypothetical. Put the falsifier dates on a calendar and decide in advance what a miss means.
Treat the incumbent G12C franchises as the short side of the thesis in your head, whether or not they're in your account — the point is that the RAS story is a *transfer* of value from the first generation to the second, not only a creation of it. And respect what the insiders did: they sold into the surge. They know things about launches that the crowd's imagination doesn't. ### The rows, in the same font as the losses *One:* **60% that Rasonque's third-quarter net sales, reported in early November, exceed RBC's $28 million.** Mechanism: more than 2,000 patients already on drug through expanded access, converting to paid therapy.
Falsifier: a slower transition than expected — the company itself has said it takes "a few months." And note the trap: a beat of $28 million can still be fined if the crowd wanted $60. *Two:* **55% that Revolution closes 2026 below Friday's $210.02.** Mechanism: a flawless launch is priced, insiders are sellers, and the next real re-rating event is a lung readout in 2027. Falsifier: a revived takeover approach, a first-line label acceleration, or lung data pulled forward. This is a row about a price, not about a company.
The company is extraordinary either way. *Three:* **65% that Roche's divarasib is FDA-approved by the end of 2027.** Mechanism: a positive head-to-head Phase 3 on PFS and OS against approved comparators, with filings planned. Falsifier: an advisory-committee surprise over the comparator design. *Four:* **60% that combined Lumakras and Krazati sales are lower year-over-year in the fourth quarter of 2027.** Mechanism: divarasib and the RAS(ON) class taking the G12C patient. Falsifier: first-line chemotherapy-combination wins for sotorasib in CodeBreaK 202 and 301 that expand the incumbents' reach faster than the challengers erode it. *Watch, no odds:* the RASolve 301 lung readout in 2027; RASolute 303 in first-line pancreatic cancer; zoldonrasib's pivotal path in G12D; Astellas's degrader; Erasca's second-half data; a Merck–Revolution sequel; and the Priority Voucher program's next recipient, because a five-week clock changes every biotech's discount rate. --- ## The honest fine print This is analysis and opinion — not investment advice, and not a recommendation to buy or sell anything.
Nobody at this desk is a physician. Every clinical figure above is quoted from the trial presentation, the FDA label, or the company, and the drug's own label carries warnings for severe skin toxicity, mouth sores, gastrointestinal perforation, lung inflammation and harm to a fetus. Rasonque is not a cure.
The survival figures are medians in a second-line population with a median follow-up of 8.5 months, and longer follow-up can move them. Valuation figures and analyst estimates are as published and as of Friday's close. The Merck talks are as reported by the FT and the WSJ; neither company has characterized them publicly.
Do your own research — and if pancreatic cancer is in your family, talk to an oncologist, not a stock column. *Fact-check notes on the first draft, printed because that's the house rule:* it said Revolution "turned down" $30 billion; the reporting says the talks ended over valuation, so it now says the company walked away from talks. It dated Shokat's discovery to "April 2013"; only the year is sourced. It called the non-G12 patients in the trial "a handful"; there were about forty.
It compared Revolution's 178 percent to "a flat-ish S&P"; the S&P was up less than 13 percent at the time. It said community practices had "never handled a RAS inhibitor's rash"; they've handled the first-generation drugs, so it now says this drug's. It said every KRAS drug depends on Foundation Medicine's testing; several labs test for G12C, so it now says one of the largest franchises.
And it called China "the world's largest pancreatic-cancer market," a superlative this desk couldn't source, now removed. --- ## One image before I let you go In 2013 a chemist in San Francisco looked at a protein that had defeated the entire pharmaceutical industry for thirty years and saw a single sulfur atom out of place. Thirteen years later, a woman with metastatic pancreatic cancer whose first treatment has stopped working can take a pill and expect, at the median, to see the following autumn — twice what she could have expected in March. That is the whole story.
Everything else in this column — the $45 billion, the $477,600, the offer that never closed, the rows and their falsifiers — is the market's attempt to put a price on that sulfur atom. The market will get the price wrong in both directions, repeatedly, on schedule. The science won't move.
And thirteen percent is finally a number that can change. — Lily **Tickers in play:** RVMD · RHHBY · LLY · MRK · AMGN · BMY · ALPMY · ERAS · ELTX · ONC · XBI · IBB --- *Sources: Revolution Medicines press releases — RASolute 302 topline (Apr 13, 2026), NDA acceptance (Jul 22), Q2 2026 results (Aug 5), BeOne collaboration (Aug 10), FDA approval of RASONQUE (Aug 26, via GlobeNewswire; indication wording, no companion diagnostic, 300 mg dosing, full approval, National Priority Voucher, warnings, adverse reactions ≥20%, availability, Goldsmith and Wolpin quotes); OncLive's report of the ASCO 2026 RASolute 302 presentation (enrollment 248/252 and 228/231; OS 13.2 vs. 6.6 mo, HR 0.40, 12-month rates 53.3%/18.7%; PFS 7.3 vs. 3.5, HR 0.45; ORR 33.2% vs. 11.8%; ITT OS 13.2 vs. 6.7, HR 0.40; time to pain deterioration 9.2 vs. 3.8; grade ≥3 TRAEs 43.6% vs. 57.5%; discontinuations 1.2% vs. 11.2%; dose reductions 36.1%; one grade-5 pneumonitis; treatment duration 6.2 vs. 1.5–3.2 mo; data cutoff Feb 10, 2026; follow-up 8.5 mo; NEJM 2026;394:1790–1802 Phase 1/2 data); Newsweek, Aug 26 (list price $39,800/30 days, ≈$477,600/yr, Part D, Kasi and Diamantas quotes, 2,000+ expanded-access patients); BioPharma Dive, Aug 26 (RBC $11.5B peak, $28M Q3, $1.1B by end-2027, de Claro quote); Insider Monkey via