The third company to crack an "undruggable" cancer target just beat the two that got there first

Biotech & Longevity 5 min 80 sources

Roche's late-arriving lung-cancer drug outperformed Amgen's and Bristol Myers Squibb's in a head-to-head trial — the week's clearest lesson that being first proves a thing is possible, but rarely means you keep the prize.

Key takeaways

Roche's lung-cancer drug beat two rivals that reached the same "undruggable" target years earlier — a reminder that arriving first proves a thing is possible without guaranteeing you keep the reward.
Alzheimer's research split two ways this week: one team found why some brains resist the disease, another found how it spreads between cells in mice — both early, neither a treatment yet.
AI drug design moved from promise to signed deals worth billions, but it only speeds the earliest step; the years of testing that decide whether a drug works are untouched.

For 40 years, a mutated protein called KRAS was the cancer target everyone wanted and nobody could hit. This week, the company that reached it last beat the two that reached it first.

Roche’s late lung-cancer drug beat the pioneers

On Wednesday, Roche reported that its experimental drug divarasib beat both Amgen’s Lumakras and Bristol Myers Squibb’s Krazati in a head-to-head phase 3 trial [13]. A phase 3 trial is the big, final test before approval — here, 338 people with previously treated lung cancer took one of the three drugs, and the trial tracked how long they lived and how long the cancer stayed in check [13]. Roche’s drug won on both measures, though the company has not yet released the actual numbers [13].

All three drugs aim at the same thing: KRAS G12C, one version of a protein that, when mutated, tells cancer cells to keep growing [13]. KRAS was called “undruggable” for decades — its surface is smooth, with no obvious pocket for a drug to grab. Amgen and BMS were the ones who finally found a grip. The FDA, the US drug regulator, approved Lumakras in 2021 and Krazati in 2022 [13].

Here is the turn. Those two drugs were genuine scientific firsts against a target that had defeated everyone — and their sales disappointed [13]. Lumakras took until 2024 to pass the modest $347 million figure analysts had once expected for its first year; it reached $363 million last year [13]. BMS paid $5.8 billion to buy Krazati and booked $205 million in sales of it last year [13]. Roche, arriving years later, argued its molecule simply gripped the target harder — better potency, better selectivity in lab tests — and bet everything on proving it in a direct fight [13]. The bet paid off. Roche now forecasts peak sales of up to $2.5 billion and plans to file for approval [13].

The caveat: Roche has shared no survival numbers yet, only that it hit its goals [13]. Cross-trial hints are not the same as a clean head-to-head readout, and the full data won’t land until an upcoming medical meeting [13]. But the direction is clear — the pioneers proved KRAS could be drugged, and a follower used that proof to build something better.

The Alzheimer’s field learns why some brains don’t break

Two findings this week came at Alzheimer’s from opposite ends — one asking why some people resist it, one asking how it spreads.

Roughly 30% of older adults whose brains carry the full biological signature of Alzheimer’s never develop symptoms [19]. A team at the Netherlands Institute for Neuroscience reported evidence that the answer may lie in a rare group of immature brain cells — cells that, in resilient brains, appear to survive damage rather than die [19]. “Around 30 percent of older adults who develop Alzheimer’s disease never experience its symptoms,” said senior author Evgenia Salta. “We really don’t know why” [19]. If that resilience can be understood, it points toward protecting memory instead of only clearing the damage [19].

From the other direction, University of Utah researchers found in mice how the disease travels between brain cells [32]. A protein called Arc — normally a helper in neuron-to-neuron messaging — appears to carry toxic tau, the protein that kills brain cells, from sick neurons into healthy ones [32]. The finding suggests a new strategy: stop tau from reaching healthy cells rather than trying to erase it [32]. The published work was in mice — most things that work in mice never reach people [32].

The contrast with the money-and-mechanism news is sharp. This week a stealthy startup paid up to $2.2 billion for an Alzheimer’s asset [1], while Merck halted a mid-stage Alzheimer’s trial after its drug underwhelmed [28]. The science advances unevenly; the checkbook does not wait for it.

AI drug design stops being a slide deck

The AI-in-medicine story shifted this week from promise to placement. Anthropic, the AI company, announced it will develop drugs of its own [65], and its new science tool debuted with claims of antibiotics designed from a text prompt [76]. Around it, the deal machine ran hot: Insilico, an AI-drug-discovery firm, signed a $600 million collaboration with the drugmaker Takeda and a separate agreement worth up to $2.5 billion with SK [3][7].

The mechanism is worth stating plainly. These systems predict which molecules might bind a target, narrowing millions of candidates before a single one is made in a lab [76]. That saves time at the front of the pipeline. It does nothing to shorten the years of animal and human testing that decide whether a molecule is actually safe and works — the part where most drugs fail. A prediction is a hypothesis, not a cure.

The under-covered corner: neglected-disease vaccines gather pace

Away from the billion-dollar deals, the diseases that don’t make Western headlines had a quiet good week. An early-stage vaccine for schistosomiasis — a parasitic worm infection affecting hundreds of millions, mostly in poorer countries — showed strong, lasting immune memory in first-in-human trials [12]. Countermeasures for Lassa fever, a viral illness endemic to West Africa, are moving forward [38]. And a trial of an Ebola treatment began in the Democratic Republic of Congo [80].

None of these are approvals. Early immune signals are the softest kind of result — a vaccine that stirs the immune system in a lab test can still fail to protect anyone in the field. But these are the illnesses that rarely draw the capital the KRAS race drew, and this week they moved.

02 · Lesson · why it matters

Why the one who proves it can rarely keep it

The hardest part of any breakthrough is showing it can be done at all — and the moment you show it, the knowledge belongs to everyone chasing you.

The wall stood for forty years

KRAS was the target cancer researchers dreamed about and dreaded. It sits inside a huge share of the deadliest tumours, telling cells to keep multiplying. Everyone knew hitting it would matter. And for four decades, nobody could. The protein’s surface is smooth — no crevice, no handle, nothing for a drug to grab.

Then Amgen and Bristol Myers Squibb found a grip. Their drugs, approved in 2021 and 2022, were the first ever to drug KRAS. That is not a small thing. It was one of the genuine “impossible, then done” moments in modern medicine.

This week, a third company beat them both.

The first crack is worth more to the followers than to the finder

Here is the part worth sitting with. When Amgen and BMS proved KRAS could be drugged, they did not just make a product. They made a fact — the fact that a grip exists. And a fact, once it’s out, cannot be kept.

Before the first drug, every rival faced the same terrifying question: is this even possible? After it, that question was answered for free. Roche did not have to gamble on whether KRAS could be hit. It already knew. It could spend all its effort on the narrower problem of hitting it better — a stronger grip, a more selective molecule — because the existential risk had been retired by someone else.

The pioneer pays for the proof. The follower inherits it. That asymmetry runs through far more than cancer drugs. The first team to reach the summit, ship the product, win the case, break the record — they buy down the uncertainty for everyone behind them. “It can be done” is the expensive part, and it’s the part that leaks.

Arriving first is not the same as winning

We tend to collapse two different things into one word: first. First to prove it works, and first to win the market. They come apart constantly.

Amgen and BMS were first to prove it. Their sales were modest — one drug took years to limp past the low bar analysts had set for it. BMS paid $5.8 billion for its version and booked a fraction of that in yearly sales. Roche, arriving late, watched all this, refined the molecule, bet on a head-to-head fight, and won it. The reward did not go to the ones who opened the door. It went to the one who walked through it with a better answer.

This is not a story about the pioneers failing. Their drugs work; patients are alive because of them. It is a story about where value settles. The one who removes the doubt and the one who captures the prize are often not the same person — and the gap between them is exactly the size of what the second one learned from the first.

The arrangement nobody voted for

Notice the quiet structure under all this. There is no rule saying the finder keeps the reward — but there’s also no rule making sure they don’t lose it. The system that funds this work treats a proven-possible target as an open field. The instant KRAS was shown to be druggable, it stopped being one company’s frontier and became everyone’s opportunity.

That arrangement serves patients — three drugs competing means better ones and, eventually, cheaper ones. It also means the company that shouldered the forty-year risk has no protected claim on the fruit. Both things are true. The structure that rewards the pioneer’s courage and the structure that hands their proof to their rivals are the same structure. You cannot have the second drug without the first, and you cannot stop the first from teaching the second.

We are all standing on proofs someone else paid for

Step back from the boardrooms. Almost everything you rely on today rests on a first crack somebody else absorbed the risk for. The people who proved a thing possible are usually not the ones who got rich from it, and often not the ones you’ve heard of. The follower who did it better gets the name and the market; the finder who did it first gets a footnote and a lawsuit.

None of us can see the whole ledger — who paid for the proof we’re now standing on, whose forty years of failure cleared the ground under this week’s win. The reward and the risk sit in different hands, and the seam between them is almost invisible from any single seat. That should make anyone slower to credit the winner and quicker to wonder who cleared the path.

03 · Lab · your turn

The Pioneer's Bet

Choose when to enter a race against an impossible target, and feel how the one who proves it possible rarely keeps the reward.

04 · Hope · carry this

The reason no wall stays standing forever is that the first crack in it belongs to everyone — one team's forty years of failure becomes the next team's starting line, and the problem gets solved faster than any of them could have managed alone.