Personalised gene therapies are here & AI can help widen their availability
A few days after KJ Muldoon was born, he was diagnosed with a rare genetic disorder. Doctors told his family:
“Your son is very sick. The best place in the world to be when your child is sick is next door.”
Long story short, KJ became first patient in the world to be treated with a personalised gene therapy that was custom-designed and tested for his specific genetic variation, within just a few months.
But, what if the best place in the world to tackle KJ’s sickness wasn’t next door? Can we work towards a world where for every kid born like KJ with rare genetic disorders, treatments can be custom designed if necessary? That’s what we’ll delve on in this blog.
Biology often surprises us with stories stranger than science fiction—and sometimes, those stories become hope. In early 2025, a medical milestone quietly unfolded: a baby named KJ Muldoon became the first human to receive a personalized CRISPR therapy tailored to his exact mutations, launching a new era for precision medicine. This was not a one-size-fits-all solution. It was a therapeutic N=1—designed, tested, and administered with the singular purpose of saving one life.1
The story of KJ Muldoon is not just a scientific milestone—it’s a glimpse into a future where lifesaving therapies are no longer limited by time, trial-and-error, or geography. AI is the bridge that can transform N=1 therapies from rare exceptions into scalable solutions. By reducing design timelines, optimizing delivery systems, and predicting outcomes more accurately, AI makes it possible to extend the promise of personalized gene therapy to many more patients—especially in parts of the world where rare diseases are often overlooked. The challenge now is not just scientific—it’s about building the right ecosystems to make these breakthroughs accessible, affordable, and global.
At the heart of the innovation that helped treat KJ was BE-Hive, a machine learning-powered tool that transformed what could have been years of trial-and-error into months of predictive, data-driven design. It was a race against time, and biology, computation, and clinical medicine sprinted in unison.
The Diagnosis: CPS1 Deficiency and a Life at Risk
KJ Muldoon was diagnosed within 48 hours of birth with carbamoyl phosphate synthetase 1 (CPS1) deficiency, a rare and often fatal genetic disorder that disrupts the urea cycle, the body’s mechanism for detoxifying ammonia. Without the CPS1 enzyme, ammonia accumulates rapidly, causing lethargy, respiratory distress, neurological damage, and, without intervention, death.
Standard treatments—low-protein diets, nitrogen-scavenging drugs, and even liver transplants—carry limited success and high risk, particularly in infants. In KJ’s case, doctors found two devastating mutations in his CPS1 gene: a C-to-T transition and a G-to-T transversion; these disrupted the enzyme to be encoded. The clock was ticking, and traditional paths offered little hope.
The Breakthrough: A Custom-Built Gene Editing Therapy
Enter Dr. Rebecca Ahrens-Nicklas and her team at the CHOP. Working with collaborators from the University of Pennsylvania, they proposed an audacious plan: craft a bespoke in vivo CRISPR therapy, customized to correct KJ’s specific mutations. In just six months—a previously unthinkable timeline—they turned the concept into clinical reality.
This effort involved more than just editing DNA—it required assembling an entire precision gene-editing platform, combining wet-lab ingenuity with powerful AI-driven insights.
Precision Editing with Adenine Base Editors (ABEs)
Unlike traditional CRISPR systems that cut DNA, this therapy uses ABEs—molecular machines that can rewrite DNA one base at a time, minimizing the risk of unwanted mutations. But to find the right editor for KJ’s unique mutations, the team turned to a powerful ally: BE-Hive.
BE-Hive: The AI Brain Behind Personalized Editing
BE-Hive (Base Editing Hive) is a deep learning platform trained on data from over 38,000 gene editing experiments. It consists of two core models:
- Base Editing Efficiency Model: Uses gradient-boosted regression trees to predict how effectively a given base editor and guide RNA (gRNA) pair will work on a specific mutation in a given cell type.
- Bystander Editing Model: A deep autoregressive network that forecasts unintended edits within the editing window, helping avoid collateral damage.
BE-Hive enabled the design of a highly specific gRNA, termed “k-abe”, that steered the ABE precisely to KJ’s mutations, converting harmful T-A base pairs back into healthy G-C configurations without double-strand breaks.
From Bench to Bedside: Testing and Delivery
Once designed, the therapeutic components were packaged into lipid nanoparticles (LNPs)—tiny carriers optimized to deliver the gene editors to liver cells, where CPS1 is normally expressed. These LNPs were tested in animal models (monkeys and mice), confirming both safety and liver-targeted delivery.
At six months of age, KJ received his first infusion. Weeks later, his doctors were able to increase his protein intake—a sign that the therapy was working and the edited CSP1 was mended. Two more doses followed, and the early results were encouraging: KJ’s body was beginning to process ammonia again enabling higher intake of much needed protein in diet to grow well.
Why This Matters: A New Paradigm for Gene Therapy
This wasn’t just a treatment—it was a proof of concept for the future of N=1 personalized therapies. In rare diseases, where patient numbers are too low for conventional trials, such bespoke approaches may be the only option. Tools like BE-Hive transform this from a manual, years-long process into a months-long pipeline, reducing wet-lab burden and making treatments feasible even for ultra-rare conditions.
However, challenges remain: regulatory flexibility, ethical frameworks, manufacturing scalability, and long-term efficacy must all be addressed. But the success of KJ’s case opens a new chapter in medicine, one where AI and biology converge to rewrite even the most challenging genetic scripts.
Scaling Precision Medicine: Building Global Ecosystems for Personalized Gene Therapies
The success of this project offers an urgent call to action for countries like India, where rare diseases often go undiagnosed and untreated due to infrastructure gaps. India can be at the forefront of this revolution by investing in:
- AI-Driven Design: Organizations like ours are already developing machine learning tools for gRNA design, mRNA sequence optimization, and LNP formulation—mirroring the BE-Hive approach locally.
- Academic and Industry Partnerships: With research institutions under CSIR and ICMR, and an established bioprocess industry, India has the talent and infrastructure to build a complete pipeline from in silico design to clinical delivery.
- Regulatory Innovation: Learning from FDA’s rapid pathways for N=1 therapies, Indian regulators can design agile frameworks for ethical and expedited approval.
With the right ecosystem, Hyderabad and other biotech hubs in India could emerge as global centers for AI-powered precision therapies, bringing cures not just to KJ, but to millions whose rare conditions still lack answers.
A Note of Caution: The Other Side of Gene Therapy
While groundbreaking cases like KJ Muldoon’s highlight the transformative potential of personalized CRISPR therapies, it is essential to approach such advances with a balanced perspective. Gene therapies do not always yield the desired outcomes, and in some cases, they have resulted in serious, even fatal complications.
For example, a case, reported by Fahreddin Palaz et al. in the New England Journal of Medicine in 20242, documented the tragic outcome for a 27-year-old patient with Duchenne muscular dystrophy (DMD) who received a high-dose rAAV9-based gene therapy designed to upregulate cortical dystrophin. Within six days, the patient developed acute respiratory distress syndrome (ARDS) and cardiac dysfunction. Despite medical intervention, he passed away just eight days after treatment. A postmortem revealed severe lung damage and cardiomyopathy, likely triggered by an innate immune response to the therapy. This sobering case is a stark reminder that while gene editing is powerful, it is also complex, context-specific, and not without risk. Clinical progress must be pursued with caution, thorough validation, and ethical oversight—especially when patient lives hang in the balance.
More recently, Rocket Pharmaceuticals3 paused its pivotal Phase 2 trial of RP-A501, an AAV9-based gene therapy for Danon disease, after a patient death. The patient showed early signs of capillary leak syndrome a week after infusion, followed by an acute systemic infection that ultimately led to death. Notably, the FDA has already imposed a clinical hold before the fatality, signaling the seriousness of the adverse event. Rocket has since identified a C3 inhibitor used in the pretreatment regimen as a possible cause and is actively working with the FDA to adjust their approach. While the company remains committed to the program, the case underscores how even with robust preclinical data, unexpected complications can arise in human trials.
These cases serve as important reminders: gene therapies must be pursued with scientific rigor, careful safety evaluation, and humility. The path forward is promising, but it demands continuous learning and caution at every step.
Summary
AI is the bridge that can transform N=1 therapies from rare exceptions into scalable solutions. By reducing design timelines, optimizing delivery systems, and predicting outcomes more accurately, AI makes it possible to extend the promise of personalized gene therapy to many more patients.
Aganitha’s AI-driven approach for gene therapy
Aganitha leverages AI and machine learning to streamline critical steps in gene therapy development—from gRNA design and mRNA optimization to LNP formulation and safety evaluation. With expertise in vector engineering, payload optimization, and predictive modeling, Aganitha supports engineering of gene and cell therapies.
Interested in learning more? Contact us today. Explore more about our solutions: Gene and Cell Therapies.
Disclaimer
This blog is intended for informational and educational purposes only. Aganitha does not offer clinical services or directly administer gene therapies. Our role is focused on supporting the research and development ecosystem through AI-driven computational tools and in silico platforms. Any references to clinical cases or therapies are shared to illustrate scientific advances and do not imply that Aganitha is involved in patient care or clinical decision-making.
- Musunuru, K., Grandinette, S.A., Wang, X., et al. (2025). Patient-Specific In Vivo Gene Editing to Treat a Rare Genetic Disease. The New England Journal of Medicine. DOI: 10.1056/NEJMoa2504747 ↩︎
- Palaz, F. et al. (2024). Death after High-Dose rAAV9 Gene Therapy in a Patient with Duchenne’s Muscular Dystrophy. New England Journal of Medicine. DOI: 10.1056/NEJMoa2307798 ↩︎
- Taylor, N. P. (2025, May 27). Rocket crashes as gene therapy patient dies, FDA imposes hold. Fierce Biotech. https://www.fiercebiotech.com/biotech/rocket-crashes-gene-therapy-patient-dies-fda-imposes-hold ↩︎