there are two ways of telling health innovation, the first consists of stacking promises with an AI that will see everything, predict everything, cure everything, and the second, which begins with a more uncomfortable observation, which is that despite decades of efforts, certain cancers remain therapeutic dead ends, with treatment lines that barely move, late diagnoses, and a clinical failure rate that makes R&D a brutal statistical game.
Cure51 chooses a third path, starting from exceptions. Its CEO, Nicolas Wolikow, assumes a methodological reversal which wants to study exceptional survivors of cancers known to be among the most aggressive, collect their clinical data, their images, and especially their tissues, then use this material to go back up the chain, isolate survival mechanisms and, if possible, transform these signals into therapeutic targets.
If the idea is simple to state, it is much more difficult to execute, and to talk about it we invited Nicolas Wolikow, the co-founder alongside Simon Istolainen of CURE 51 in Perspectives, our show which decodes innovation around the world, in partnership with Canalchat Grandialogue
The end as a starting point: the outlier becomes a protocol
The heart of the project lies in a formula, starting from the end, where a large part of clinical research observes cohorts progressing towards failure, Cure51 is interested in patients diagnosed late, in the advanced phase, on cancers where survival is often measured in months. And in the middle of this statistic, some abnormal trajectories with one to two percent of survivors, sometimes after years, against all expectations.
Rarity becomes the initial filter, CURE51 does not start from a gigantic dataset that we hope to make meaningful, but starts from an extremely rare event that we hope to make explainable. For this reversal to have meaning, we need discipline and not confuse individual story and scientific proof. Cure51 insists on a retrospective approach already explored in certain academic works (in the United States, Australia, France via Unicancer).
The real obstacle is not the algorithm: it is access to hospitals
In healthcare, innovation rarely comes up against a shortage of ideas; clinical data is not a self-service resource. To develop its project, CURE51 deployed a very structured methodology, which begins by identifying the relevant establishments, locating the “principal investigators” (the lead researchers), convincing the scientific and ethical committees, contracting data transfers, then verifying, case by case, that the archive really exists.
AI: accelerate cycles, then aim for prediction
For Nicolas Wolikow, AI is less a magic wand than an acceleration engine. On clinical data and imaging, part of the progress is due to matching, clustering and categorization capabilities, sometimes closer to classic machine learning than to the grand narrative of “intelligence”.
The breakthrough, according to him, is taking place in biology with multi-omics analysis, single-cell resolution, understanding of the tumor micro-environment, and above all the progressive transition from observation to prediction. After a first phase where we learn to read, a second phase begins, where we hope to anticipate.
Europe, United States, China: the advantage is not only regulatory
We took advantage of the exchange to address the question of global competition, it offers us an interpretation in three poles: the United States as the economic center of gravity of pharma, China as an industrial and scientific accelerator in certain segments, France and Europe as a space of talents… but of infrastructural dependence. Europe therefore suffers from the incompleteness of the chain: cloud, hosting, certain analysis technologies, certain hardware layers, which remain dominated by American service providers.
A two-stage economic model: quick proof, long value
The first, strategic, stage is co-development, namely identifying targets, building candidates, then partnering with laboratories capable of absorbing the burden of clinical development, via milestones and royalties.
The second level is a more immediate revenue ramp: using the base as a tool for evaluating external assets. A pharma submits molecules; Cure51 returns a score based on comparison with its exceptional survivors, to guide a “go/kill” decision. This product has one virtue: it transforms the data asset into a salable service without giving up access to the raw data, while building the trust that will later make a more structuring IP deal possible.
In 2024, Cure51 raised €15 million in funding during a seed round led by Sofinnova Partners, Hitachi Ventures GmbH, Life Extension Ventures, Xavier Niel and Olivier Pomel, CEO and co-founder of Datadog. The startup is preparing to achieve a series A
