Did AI cure cancer? Moderna surged 177%, but the real breakthrough is more important than the headline

The bottom line

The title "Humanity uses AI to cure cancer for the first time" is too much.

But the medical breakthrough it borrowed from was indeed big enough: Moderna and Merck announced that the personalized mRNA neoantigen therapy intismeran autogene combined with Keytruda achieved the two endpoints of recurrence-free survival and distant metastasis-free survival in a Phase 3 trial involving 1,137 high-risk melanoma patients.

This is the first personalized neoantigen therapy to achieve positive Phase 3 results and the first mRNA cancer therapy to report positive results in Phase 3. After the news was announced, Moderna's stock price rose by about 177% during the session.

But there are three boundaries that cannot be crossed:

• The trial measures adjuvant treatment after complete surgical removal of melanoma, not the elimination of all advanced tumors with one shot.
• The two companies have only announced top-line results so far. The specific risk reduction in Phase 3 has not yet been disclosed. The overall survival period is still under observation.
• intismeran is still an unapproved investigational therapy; "curing cancer" is not what this trial has proven.

What is truly worth writing into history is not an exaggerated title, but a process that once resembled laboratory science fiction, passing through the Phase 3 "Valley of Death" for the first time: Everyone's tumor is different, so a drug is designed and produced individually for each person.

The market went crazy first: A top-line announcement pushed Moderna up about 177%

On August 19, Moderna and Merck announced that the INTerpath-001 trial met the primary endpoint of RFS and also met the key secondary endpoint of DMFS.

Translated into human words: When adding personalized mRNA therapy, patients relapsed later after surgery and developed distant metastases later than Keytruda alone, and the difference was both statistically significant and clinically meaningful.

Wall Street reacted faster than a medical conference. AP reports that Moderna shares have "more than doubled," and Business Insider and The Times put the intraday figure at about 177%. Investors are betting not just on one melanoma indication, but on whether this "one person, one drug" platform can be replicated to target more tumors.

However, the skyrocketing stock price proves that the market has changed its expectations, which does not mean that the drug has been approved, nor does it mean that all cancers have antidotes.

This drug isn’t something you pick up off the shelf: it’s made to fit your tumor

The intismeran idea is counterintuitive.

Traditional medicines try to give the same ingredient to as many patients as possible; intismeran starts from the patient's tumor tissue and blood, reads DNA and mRNA sequencing data, and finds the unique mutation "fingerprint" of the tumor.

It then needs to single out the neoantigens among a large pool of candidate mutations that are most likely to elicit a response from the immune system. Each therapy will ultimately encode up to 34 neoantigens, packaged into a customized mRNA therapy for this patient.

Once inside the body, the neoantigen sequences encoded by the mRNA are translated and presented to the immune system, with the goal of training T cells to recognize cancer cells with the same mutations. Keytruda is responsible for lifting the immune "brake" of PD-1, giving immune cells that have recognized the target a greater chance to launch an attack.

To use a metaphor that’s not entirely rigorous but easy to understand: Keytruda is taking the brakes off, while personalized mRNA therapy is trying to teach the immune system “what this patient’s cancer cells look like.”

AI is really involved, but it’s not like “a big model invents a miracle drug in a few hours”

“AI is just a hot topic” is also inaccurate. Moderna publicly stated in 2023 that this personalized therapy uses a series of integrated AI algorithms: They read next-generation sequencing data of tumors and blood, review genetic mutations, and predict up to 34 neoantigens most likely to trigger an immune response.

AI is also involved in something even less cool, but equally decisive in whether a therapy gets off the ground: scheduling production for large patient batches, coordinating clinical, quality, manufacturing, shipping and administration dates. Everyone’s formula is different. If design and production scheduling cannot be automated, it will be difficult to expand “one person, one drug” to thousands of subjects.

However, in the existing first-hand information, there is no evidence that "the large AI model frantically deduced protein folding within a few hours and invented the drug by itself." What is officially described is a neoantigen prediction algorithm and manufacturing scheduling system, not a ChatGPT-style big model that solves cancer overnight.

AI is an important component in this factory and design assembly line, not a doctor on stage alone.

What did the 1,137-person Phase 3 trial actually show?

INTerpath-001 was a randomized, double-blind, placebo- and active-controlled global Phase 3 trial enrolling 1,137 patients with completely resected stage IIB to IV cutaneous melanoma. Subjects were divided into 2:1 groups:

• One group received intismeran plus Keytruda.
• The other group received Keytruda plus a placebo.

Both key endpoints were met, indicating that personalized mRNA therapy brings additional benefits to standard adjuvant immunotherapy and is not just a beautiful signal in a small early-stage trial.

Officials also stated that the safety profile of this trial was consistent with previous combination therapy studies, and no new safety signals were observed. This sentence cannot be rewritten as "the side effects are similar to those of ordinary vaccines": Keytruda itself is an anti-cancer drug that may cause severe immune-related adverse reactions, and complete safety data still needs to be disclosed at academic conferences.

The most easily misused numbers: 49% and 59% are not results from this Phase 3 trial

There are already articles online that directly follow the Phase 3 announcement with "the risk of recurrence or death is reduced by 49%" and "the risk of distant metastasis or death is reduced by 59%."

These two figures are the five-year follow-up results of the earlier phase IIb KEYNOTE-942, not the Phase 3 effect size just announced this time for INTerpath-001. The Phase 3 announcement only said that the difference was significant and clinically meaningful, and that detailed risk ratios, absolute risk differences, and group data would not be presented until a later international medical meeting.

This distinction is important. The beautiful numbers in Phase II provide confidence, and the top-line success in Phase 3 provides high-level confirmation; but the two data cannot be pieced together into a conclusion that goes beyond the official announcement.

Why this could still be a platform-level turning point

The most difficult part of personalized cancer therapy is not just finding a good target, but whether it can be completed repeatedly in a real medical system: sampling, sequencing, algorithm selection, sequence design, production, quality inspection, transportation, and on-time administration.

Making a drug candidate is hard; producing 1,137 different sets of codes for 1,137 subjects and bringing them into a double-blind global Phase 3 is another level of engineering difficulty.

If complete data and regulatory review continue to support it, this success will not only prove that "melanoma can have an additional adjuvant treatment", but may also be the first time that a personalized drug design and production platform has stood up in a large confirmatory trial.

This is less suitable for a slogan than "AI cures cancer", but closer to a real scientific breakthrough.

Our judgment: remember today’s result—but not as “cancer has been conquered”

The best way to write this news is not to turn it into a cold announcement from a pharmaceutical company, nor to shout "Mankind will defeat cancer from now on."

Its real drama is strong enough:

In the past, pharmaceutical companies produced the same medicine for one disease.

Now, a platform that includes AI neoantigen selection and digital manufacturing is trying to produce a different drug for each person, and has obtained positive Phase 3 results for the first time.

What we can say today is this: Personalized mRNA cancer therapy crosses the critical threshold of Phase 3 for the first time, with AI involved in the drug design and manufacturing process.

What we cannot say today is: AI has cured cancer, or that this regimen has been proven to cure all solid tumors.

Being able to keep these two sentences in mind at the same time is the greatest respect for this breakthrough.

What to watch

• Hazard ratios, absolute benefits, stage groupings and specific adverse events in complete Phase 3 data.
• Whether overall survival ultimately shows benefit.
• When regulatory authorities accept submissions, and labeling, pricing and lead times after approval.
• Whether this platform can be repeated successfully in Phase 3 trials in other tumors such as lung cancer.
• Whether AI methods for selecting neoantigens can continue to improve as clinical data accumulate, and how this will be regulated.

FAQ

Has AI really cured cancer?

No. This time, personalized mRNA therapy combined with Keytruda achieved the endpoints of no recurrence and no distant metastasis in a Phase 3 trial of high-risk melanoma after surgery. It has not yet been approved, overall survival is still being observed, and it cannot be called "AI cures cancer."

What does AI do in intismeran?

Moderna publicly stated that an integrated AI algorithm reads tumor and blood sequencing data to predict up to 34 new antigens that are most likely to trigger an immune response; an AI scheduling system is also used to coordinate personalized production batches for each patient.

Are the 49% and 59% reductions Phase 3 results?

no. Both numbers come from the five-year follow-up of an earlier Phase IIb trial. This Phase 3 announcement has not yet disclosed specific effect sizes, only confirming significant improvements in RFS and DMFS.

Is this a vaccine to prevent cancer?

It is not a vaccine that prevents cancer in healthy people in the traditional sense. It is a therapeutic neoantigen therapy customized based on tumor mutations in diagnosed patients. This trial is used for adjuvant treatment after surgery.

Can patients use it now?

It cannot be used as a standard treatment on the market. Intismeran remains an investigational product, and the two companies said they will share data with regulatory agencies and communicate about filings.