This week, key additions to ProQR's team were announced to help drive that work: Chris Hart, PhD, as Chief Data and AI Officer; Thomas Wolf, PhD, joining the Board of Directors as strategic AI advisor; additionally, Gerard van Westen, Professor of Artificial Intelligence and Medicinal Chemistry at Leiden University is joining ProQR's AI team alongside his academic position. In this post, ProQR and the new ProQRians lay out that strategy in plain terms: what's actually changing, and why it matters well beyond the molecules AI helped discover.
ProQR's Next Chapter in Creating Medicines for Patients Why AI, Why Now
Written by
ProQR Therapeutics
AI now sits at the center of how ProQR pursues its mission: discovering, developing, and delivering transformative RNA-editing medicines to patients with high unmet medical needs, faster and with a higher chance of success. Getting that right takes more than a new algorithm layered onto old workflows - it means rethinking how discovery and development teams work together across the company, it means rethinking how and what data we generate to answer out scientific and development questions, end to end.
Building on strong proof points
Before getting into the strategy, it's worth being clear about where it stands today. ProQR isn't announcing an intention to experiment with AI - it already has a molecule to show for it: AX-0811, the company's first AI-discovered editing oligonucleotide (EON - the class of molecule Axiomer uses to make a precise, single-letter RNA edit), is already in clinical testing, with initial target-engagement data expected in early January. The same AI-driven design approach is used across programs, with the second program shaped by this approach, AX-2911, expected to follow into the clinic in the coming months.
That puts ProQR in a leading position: the only company with multiple RNA-editing therapies already in clinical trials, with two more expected to enter the clinic within the next several months. AI is one of the reasons that lead is widening rather than closing.
We've been building AI into Axiomer for a while now - quietly, because we wanted to know it worked before we talked about it. AX-0811 exists because of that work. It's already in a clinical trial, not just on a slide. That's the proof point everything else here builds on.
The progress made with AI integration in our drug discovery process has led to tremendously improved molecules that outperform the EONs we identified through the manual process. Leveraging roboticized data generation to rapidly feed the models is expected to drive further innovation and advancements.
Why RNA editing is a particularly good fit for AI
The EON design space is large: sequence and chemical choices interact, and together determine editing activity, selectivity, and potency. AI and machine learning help map those relationships and search the space systematically, identifying the combinations most worth testing.
ProQR has ten years of proprietary data to train on, and that dataset has proved highly valuable. It gives the models an experimentally grounded view of how EON design choices translate into editing performance. That foundation enabled one of ProQR’s first AI-driven successes: using the model to identify an EON with substantially improved performance over the manually designed candidates.
The real value of an AI model is not simply that it makes predictions. It is that it helps us see structures in the data and therefore what to investigate next. The experiments then test those ideas, maximize information gain, and improve the model in return. That creates a more deliberate and steered way to learn from biology.
The Platform Opportunity at ProQR
What drew Dr. Hart, Dr. Wolf and Dr. Van Westen to ProQR the most, is that Axiomer is a platform, not a single-program discovery engine. The same underlying biology and design space recur across targets, so what ProQR learns about target selection, EON design, assays, and development can carry forward. Each program adds new evidence about what works and what does not. That is the source of the platform’s compounding value.
The lab-in-the-loop makes this learning operational. It represents an iterative cycle: AI models propose EON designs, lab experiments test them, and the results feed back into the models, so each round of experiments makes the next round of designs better. The process itself is continuously improved by adding more relevant assays to the HTS pipeline, increasing assay throughput, shortening design-test cycles, experimenting with new chemistries, and improving both the models and the designs they generate. The result is not a marginal improvement to screening; it is a different discovery paradigm. In current programs, ProQR can optimize EONs for potency and stability while evaluating splicing effects and toxicity in parallel, early enough to redesign around those liabilities. Instead of discovering late whether a candidate has the right profile, the platform helps shape that profile from the start, materially reducing avoidable clinical risk.
Across the broader development cycle, AI provides the connecting layer, linking target biology and discovery data with preclinical and, increasingly, clinical evidence. As clinical data accumulates, it will help reveal which biological relationships and assay signals genuinely predict outcomes in patients. This way, ProQR is building towards a system where every outcome feeds back into the models, strengthens the platform, and improves the programs that follow.
AI is most powerful in drug discovery when it creates a feedback loop across the stages, not when it optimizes each stage in isolation. The question is how to feed information backward: use experimental results to improve the next designs, preclinical findings to improve the assays, and clinical outcomes to refine the biological hypotheses behind the next programs. Building that loop around Axiomer is the opportunity at ProQR.
Extending AI into the clinic
Clinical development is where the value of the entire platform is ultimately tested. Clinical data is scarce and expensive, but it is also uniquely informative: it shows whether the biological relationships, EON properties, and assay signals learned in discovery actually translate into patient benefit. Proper implementation of AI in this phase is more important than in any other phase, as it directly translates to the success in trials and getting a drug to market.
AI can improve the probability and efficiency of success by predicting which patients are most likely to respond, informing which endpoints will detect a real effect, and supporting trial designs that incorporate real-world data to answer the clinical question with fewer patients, in less time.
ProQR expects up to five clinical readouts across four programs over the next twelve months. AI-supported development work is designed to raise the probability of success and the speed of each of those readouts - because the mission isn't served by a faster discovery engine alone, it's served by more, and better, medicines actually reaching patients.
This is applied statistics and pattern recognition against real-world and trial data. And it compounds with everything upstream. If discovery gives you a better molecule and development gives you a smarter trial around it, your probability of success moves twice.
AI isn't a side project running in parallel to our clinical pipeline. It's becoming part of how we design and run the trials our whole pipeline depends on - which is exactly where it needs to be if the goal is getting more medicines to patients, faster.
Getting to human data faster: an integrated, bench-to-bedside platform strategy
To increase access to clinical data from its programs and derisk development earlier, ProQR is establishing a pathway for investigator-initiated trials (IITs). An IIT is an academic clinical study run in partnership with outside physicians and academic centers; using the IIT pathway in China in particular lets the company generate real, conviction-building data in patients at a fraction of the cost and time of a typical company-sponsored trial.
That speed matters for the same reason AI does: it shortens the loop between an idea and the evidence that tells you whether it's working. AX-0811 shows part of that plan already: after its ongoing European healthy-volunteer trial, it is expected to move into an IIT in paediatric biliary atresia patients. AX-2911 points at where this is heading in full: an AI-generated EON moving more directly into an IIT in China as the first-in-human study to generate patient data in the disease context. Each program does double duty - it tells ProQR quickly which molecules are the clear winners worth pushing hard, and it feeds real clinical outcomes back into the models driving the next round of discovery and development.
That matters at a bigger scale than any single program. Axiomer's approach is, in principle, applicable across hundreds of genetic disease targets where a precise RNA edit could correct the underlying defect. Capturing that opportunity efficiently is exactly why discovery, data generation, and clinical learning need to work as one connected system rather than being rebuilt by hand for every new target.
Put together, these pieces work as one machine built to capture the full value of the platform. AI-assisted EON design, paired with robotized data generation through the Ginkgo partnership, generates strong candidates quickly. IITs then generate affordable, at-scale, conviction-building patient data that let ProQR pick winners and derisk development early. And AI applied to trial design raises the probability of success of every program that follows from there. Each piece feeds the next, all pointed at the same outcome: more, and better, medicines reaching patients, faster.
The team building it
Chris Hart spent fifteen years in AI/ML-driven oligonucleotide drug development - most recently leading data science and AI/ML for Eli Lilly's Genetic Medicines team, and before that as founding CEO of Creyon Bio. “Frontier AI expertise, academic rigor, and inside-the-pipeline execution, all pulling on the same thread at the same time - that combination is what made this role attractive to me,” he says.
Thomas Wolf, co-founder and Chief Science Officer of Hugging Face - the open-source AI platform Nvidia recently agreed to acquire for $12.9 billion - joins ProQR's Board as strategic AI advisor. Over the last decade he has built one of AI's largest and most closely watched success stories, which gives him a view of most of what the AI landscape has to offer, in and out of biotech - and he doesn't join things lightly. Beyond advising the Board on strategy, he's working directly with Chris Hart's team on how the company structures its data, models, and infrastructure, bringing one of applied AI's most credible builders into direct contact with Axiomer.
I've spent a decade watching what AI can and can't do, across a lot of industries. I’m very inspired about using AI to actually help people's health. ProQR had already done tremendous work on the approach: AX-0811 in the clinic, a Ginkgo partnership generating data at a scale, and editing RNA with computational tools. I’m happy to join the ProQR board and be close enough to follow, learn and advise with real data and impactful models alongside Chris's team.
Gerard van Westen has spent his academic career researching machine learning methods for predicting how drug candidates bind and how mutations influence this process in the body - including more recently generative approaches to bespoke molecule design to target these mutations. Joining ProQR lets him keep that research going while working against real pipeline constraints instead of benchmark datasets - the sharper, harder questions that come up when a model must be right about an actual clinical candidate.
We've carefully thought through how to build the complementary AI leadership to further leverage the opportunity we have on hand. Between Chris’ 15 years of AI/ML experience in oligonucleotide optimization, Thomas’ leadership building the Frontier-AI open source platform, and Gerard’s academic leadership in AI & computational drug discovery we have a star team of leaders helping us shape how to maximize the impact on achieving our mission.
What to expect from Axiomer x AI
Near term: initial target-engagement data from AX-0811 in early January, followed by its move into an investigator-initiated trial in biliary atresia patients. Across the rest of the pipeline, expect a busy year of readouts, including AX-2911 moving an AI-generated candidate directly into an IIT in China to generate patient data quickly. Beyond that ProQR will ramp up the use of AI in target hunting, multi-objective EON design models, and all elements of clinical trial execution - from patient stratification to endpoints selection and regulatory document formation - all to shorten development timelines and increase probability-of-success. ProQR also expects to contribute some of what it learns about AI-driven RNA editing back to the field through publications, rather than treating all of it as proprietary.
We hold ourselves accountable to drive the scientific opportunity we have with Axiomer to maximal benefit for patients that are waiting for new medicines - I believe this AI strategy solidifies our efforts to achieve that.
This post reflects the current views and expectations of ProQR's management and advisors as of the date of publication and may include forward-looking statements about the company's AI strategy and pipeline, including anticipated clinical data timing and trial plans. Actual results may differ; see ProQR's SEC filings for a full discussion of risks.
