AI Alliances and a Wall Around China's Oncology Ambitions
Merck's partnership with Nobel-winning AI firm Isomorphic Labs signals a new phase in oncology R&D — one driven by computation rather than acquisition. Meanwhile, China's biotech surge hits a regulatory ceiling as the FDA rejects another Chinese cancer drug candidate.
The New Battleground: Algorithms Over Acquisitions
Global oncology drug development is undergoing a structural pivot — and two recent events, separated by ocean and regulation, mark the moment.
On one side, Merck (MSD) announced a partnership on April 19 with Isomorphic Labs, the Alphabet-spinoff AI company founded by Nobel laureate Demis Hassabis, whose AlphaFold breakthrough earned him the 2024 Chemistry Nobel. On the other, the FDA issued a Complete Response Letter rejecting ibonecestamab, a Chinese-developed bispecific antibody that had been touted as a future rival to Merck’s own Keytruda. Both stories point to the same truth: the era of treating foreign drug candidates as plug-and-play pipeline fillers is closing.
What replaces it is more expensive, more complex, and far less predictable — but also, for big pharma, potentially far more defensible.
Merck Bets on Computation, Not Just Collaboration
Isomorphic Labs brings something Merck cannot easily buy elsewhere: an AI drug-design engine called IsoDDE, built on the same protein-structure-prediction DNA that made AlphaFold famous. Under the deal, Isomorphic runs the early-stage target identification and molecular design, while Merck takes over candidate optimization, preclinical work, global clinical trials, and eventual commercialization.
The financial architecture matters too. Isomorphic receives upfront payments, milestone fees tied to development stages, and royalties on any approved product. But the real prize for Merck isn’t just a partnership — it’s speed. Keytruda, the world’s best-selling immuno-oncology drug, faces patent cliff pressure around 2028. Every month of delayed next-generation pipeline development is revenue at risk. By embedding an AI design engine at the front end of discovery, Merck is attempting to compress the longest, most unpredictable leg of drug development into a computational loop.
Sara Skarratt, Isomorphic’s chief scientific officer, framed the goal bluntly: tackle cancer targets that traditional methods simply could not approach. George Yancopoulos, Merck Research Laboratories’ senior vice president, called it a marriage of “vast clinical data” with “advanced computing models.” Translation: Merck is no longer competing only on biology. It is competing on compute.
Iboneycestamab and the End of the Easy Chinadollar
If Merck is looking forward, the FDA’s rejection of ibonecestamab is a reminder that the backward-looking strategy — buying finished candidates from fast-moving Chinese biotechs — has grown far riskier.
Ibonecestamab, developed by Hangzhou Akeso Biopharma and licensed to US-based Summit Therapeutics for North American rights, was one of the most anticipated Chinese-origin oncology assets to reach US regulatory submission. It hit a novel dual mechanism: activating the immune system while simultaneously blocking tumor angiogenesis — a pathway some saw as a direct challenge to Keytruda’s PD-1 dominance.
The FDA said no. Its objection was specific and unforgiving: the pivotal global clinical trial failed to show a statistically significant improvement in overall survival compared with the control arm. Progression-free survival improved, yes. But OS remains the hard endpoint for oncology approvals, and Merck’s Keytruda comparisons will likely reference it as well. Summit argued that patient enrollment delays meant insufficient follow-up time, and it submitted supplementary long-term data — but the FDA held.
This is not an isolated event. Across multiple Chinese biotech submissions over the past two years, FDA reviewers have increasingly demanded higher bars for survival evidence, more granular statistical justification, and more rigorous manufacturing transparency. The days when a promising Phase 2 readout and a clean mechanism were enough to secure US entry are fading.
Who Wins, Who Loses
The winners here are companies that combine computational depth with clinical-scale execution. Isomorphic Labs gains access to Merck’s $13 billion-plus annual R&D budget, decades of clinical infrastructure, and a regulated commercialization machine. Merck gains a technology platform that could accelerate candidate generation from years to months — if the AI models deliver.
Chinese biotech firms that relied on the traditional route — rapid clinical iteration, attractive early data, and licensing deals with Western partners — face a steeper climb. Akeso, which positioned ibonecestamab as a banner Chinese asset, now must explain why its strongest candidate failed to meet the FDA’s survival threshold. Other firms watching closely include InnoCare, BeiGene, and Hansoh Pharma, all of which have ambitions in the US oncology market and are tracking the regulatory tone shift.
Summit Therapeutics loses a headline candidate but retains the option to continue developing ibonecestamab, possibly in combinations or later lines of therapy. Bob Duigan, Summit’s co-CEO, insisted the company remains committed. Whether that translates into a viable second application remains unclear — especially with Keytruda-vs-ibonecestamab head-to-head data expected later this year, which could prove decisive regardless of regulatory posture.
What Comes Next
Three dynamics will define the next round of oncology competition.
First, the Merck-Isomorphic model will attract imitators. Novartis, Pfizer, and AstraZeneca have all explored or expanded AI drug-discovery collaborations. The question is whether those programs can match the integration depth Merck is pursuing — algorithm at the front end, not a bolt-on analytics service.
Second, Chinese biotech firms will have to adapt or retreat. The ibonecestamab rejection is a signal that mechanism innovation and speed are necessary but insufficient. Firms that want US approval will need to design trials with OS power from day one, invest earlier in regulatory alignment, and likely build more internal capability rather than relying on license-out exits.
Third, the Keytruda comparison trial — pitting Keytruda directly against ibonecestamab in chemotherapy-naive NSCLC patients — will carry outsized weight. If ibonecestamab survives on its own merits, the AI-driven Merck strategy looks like hedging. If it falls short, the case for computational-first drug discovery strengthens further, and the window for Chinese-origin US launches narrows even more.
The global oncology ecosystem is not being reshaped by one technology or one market. It is being redrawn at the intersection of them — where algorithms meet clinical endpoints, and where the old pipeline logic no longer guarantees passage.