At Lumanity’s Cancer Progress 2026, the goal wasn’t just to showcase innovation — it was to pressure-test it.
Held on April 9th at the New York Genome Center, the one-day meeting brought together leaders across biotech, pharma, and research to do what the event has done for nearly 40 years: create space for frank, often uncomfortable conversations about what’s actually working in oncology — and what isn’t.
By the time the final panel of the day kicked off — “Beyond Next-Gen: How Should We Engineer Future Breakthroughs?” — the tone was noticeably honest.
The panelists for the day’s final discussion were:
- Joe Guidi, PhD, VP, Worldwide Head of Immunology, Medical Affairs, Bristol Myers Squibb
- Zhen Su, MD, MBA, CEO, Marengo Therapeutics
- Axel Hoos, MD, PhD, CEO & Founder, Stealth Biotech
- Alicia Zhou, PhD, CEO, Cancer Research Institute (CRI)
- Dennis Chang, PhD, SVP, Strategy Consulting, Lumanity (Moderator)
The well-attended discussion started with a line that landed because it was true: we’ve made a lot of progress in cancer by “throwing spaghetti at the wall.”
And for a while, the panelists all agreed, that has worked in the past.
The End of Productive Chaos
Many of immuno-oncology’s biggest breakthroughs haven’t come from perfectly engineered systems. They emerged from strong biology, imperfect models, and a willingness to test hypotheses before everything was fully understood.
But as the panel made clear, that approach is starting to become outdated.
Dr. Alicia Zhou, CRI’s Chief Executive Officer, didn’t dismiss experimentation — rather, she contextualized how and when researchers should consider al dente discovery.
“There has to be the right time in the development pipeline — when I do think ‘spaghetti’ could be the right technique. But, I think when it comes to combinations, when you’re thinking about the multiple permutations that you could possibly have — that’s where we have to be more directed,” Dr. Zhou explained.
All four panelists agreed that one issue that researchers currently face is the over-indexing of experimentation – especially when it comes to combinations.
If two therapies show modest benefit, combine them. If that combination works somewhere, expand it everywhere. Repeat that across tumor types, lines of therapy, and modalities, and you get a system that produces more trials than insight.
It’s the difference between testing a recipe and dumping the whole pantry into the pot. What starts as thoughtful experimentation quickly turns into excess — more ingredients, more combinations, more variables than anyone can realistically make sense of.
Eventually, the signal gets buried under the sauce, and what began as exploration has now become an overcrowded plate, if you will.
Treating Cancer Isn’t a One-Size-Fits-All Approach
One of the most persistent challenges in oncology isn’t a lack of innovation — it’s a mismatch between the complexity of the disease and the simplicity of how we often try to treat it.
Cancer isn’t one problem waiting for a better solution. It’s a collection of fundamentally different problems, too often approached with the same playbook.
The reason why we’re not seeing great outcomes across all tumor types is different. There are very different problem sets to be solved.
She went on to unpack what “failure” actually means in oncology — and how misleading that single word can be.
A therapy might fail because it never hits the right target. Or because the tumor lacks the right biomarker. Or because it works — until the cancer adapts and escapes. In other cases, progression slows, only for metastasis or clonal expansion to take over.
And yet, those distinctions don’t always translate into how therapies are developed or tested. The mismatch between the complexity of the problem and the uniformity of the response is part of what’s slowing real progress.
AI Won’t Save Us — But It Might Help Us Think
If there was one theme carried throughout the entire conversation, it was the role of artificial intelligence, or AI, in creating new therapies and bringing those treatments to patients.
Dr. Zhou boiled it down to this: AI is useful for oncology, though it may be fundamentally limited.
“I believe generative AI is going to hit a wall. It cannot predict things that we cannot actually validate biologically in the physical world.”
Dr. Zhou cautioned the panel and those in attendance not to overestimate the fundamental understanding that AI has of how the human body biologically functions.
“I think letting AI run amok in that space is actually not useful, because there’s no way for us to validate if the prediction is a hallucination or if it’s biologically sound. Ultimately, if we can supply the model with the right underlying data, I think that’s where AI can have a real, transformative, and accelerating approach,” she said.
Dr. Zhou pointed to one of CRI’s newest endeavours, the CRI Discovery Engine, as a potential solution for the problems currently facing scientists and researchers navigating how to integrate AI into their day-to-day workflows.
The goal here is to say, can we start to really understand the mechanism of what’s actually happening?
It’s not about chasing every molecular detail. It’s about understanding the contours of the immune system — where it holds, where it breaks, and where intervention is actually possible.
Without that, even the most advanced tools are working in the dark. In other words, AI isn’t the breakthrough. It’s the tool that might finally help us see where breakthroughs are hiding.
How Immunotherapy Exposed the Knowledge Gap
Nowhere is cancer’s complexity more apparent than in immuno-oncology. Unlike targeted therapies, immunotherapy isn’t just about hitting a tumor — it’s about activating an entire system.
“You’re actually trying to start a conversation between two different cell types and systems at the same time. That’s actually very complicated,” Dr. Zhou explained.
When it fails, it’s not always clear why. Did the drug reach the immune system? Did the immune system respond? Did that response make it to the tumor — or get lost along the way?
Instead of answering those questions, the field has often responded by adding more variables — more drugs, more combinations, more trials. Activity increases, but understanding doesn’t always follow.
The consensus of the panel was clear: we’ve underinvested in truly understanding the biology of these interactions — and that gap is becoming the bottleneck.
Better Science, Broken Economics
Zooming out, the conversation turned from what’s happening in the pan to a bigger problem with the recipe itself.
The science is getting sharper. Therapies are more targeted. Outcomes, in some cases, are improving. But the system around it hasn’t kept pace. Smaller patient populations paired with billion-dollar development pathways create a structural challenge — one that scientific progress alone can’t solve.
Which raises a bigger question: if science is evolving, why hasn’t the model?
The panel pushed toward system-level changes — rethinking trial design, exploring synthetic controls, and reconsidering how therapies are approved.
These aren’t incremental tweaks. They’re fundamental shifts in how oncology operates.
Setting the Timer
Since the inaugural Cancer Progress in 1989, the gathering has always been about looking ahead — but this year, the horizon felt closer.
Dr. Zhou didn’t frame disruption as a possibility. She framed it as inevitable.
AI is going to fundamentally transform the way we do everything.
And AI is only part of the story.
Global competition is accelerating. Blockbuster drugs are approaching patent cliffs. New players — and new models — are entering the ecosystem.
The pace of change is no longer theoretical. It’s already underway.
The “Take Home” Portion
If Lumanity’s Cancer Progress is designed to challenge assumptions, this panel delivered.
Not by rejecting how the industry got here — but by making it clear that those same approaches won’t take it forward.
Experimentation still matters. Sheer luck still plays a role at times. But the next wave of breakthroughs won’t come from throwing more of anything at the wall.
Discoveries will reveal themselves as a more intentional approach is applied, because at this point, the question isn’t whether we can generate more ideas – it’s whether we can finally make sense of the ones we already have.
