Origin Bio Is Programming DNA Switches to Make Gene Therapy Actually Safe

The Macro: Gene Therapy Has a Precision Problem

Gene therapy is one of the most promising areas of medicine. The idea is straightforward: instead of treating symptoms with drugs, you fix the underlying genetic problem by delivering the right DNA to the right cells. In practice, it is far more complicated than that pitch implies, and one of the biggest challenges is control.

When you deliver a therapeutic gene into a patient’s cells, you need it to turn on in the right cells, at the right time, at the right level. Too much expression and you risk toxicity. Too little and the therapy does not work. Expression in the wrong cell type and you get off-target effects that can cause serious harm. The gene is just the payload. The regulatory DNA, the enhancers and promoters that control when and where that gene activates, is what determines whether the therapy is safe and effective.

This is the part that most people outside of biotech do not realize. The therapeutic gene itself is often the easy part. The hard part is the on/off switch.

Current approaches to regulatory DNA design rely heavily on natural sequences pulled from existing genomes or from viral vectors like AAV (adeno-associated virus). These work to some extent, but they are blunt instruments. Natural promoters were not designed for therapeutic precision. They evolved for other purposes entirely. The result is that many gene therapies have limited cell-type specificity and imprecise expression levels, which contributes to the safety concerns that have slowed the field.

Origin Bio, backed by Y Combinator (W25), is approaching this problem with AI. They design novel synthetic regulatory DNA sequences, enhancers and promoters that do not exist in nature, optimized to program precise gene expression patterns in specific cell types.

The Micro: Synthetic DNA Switches Designed by AI

Yash Rathod (CEO) and Malhar Bhide (CTO) founded Origin with the thesis that AI can design regulatory DNA sequences that outperform anything found in nature. Their team includes people from NVIDIA, UC Berkeley, UPenn, and UIUC, with scientific advisors from MIT and UCSF.

The recent launch of their “Switch” platform makes 10,000 AI-designed regulatory DNA sequences available for research purposes. These are not random sequences. Each one is designed and predicted by their Axis model to drive specific expression patterns in specific cell types. The scale matters. Having 10,000 sequences to choose from, rather than the handful of well-characterized natural promoters that most gene therapy programs rely on, opens up a much larger design space.

The competitive field includes Dyno Therapeutics, which uses AI to engineer AAV capsids (the delivery vehicle for gene therapies, not the regulatory elements). There is also Generate Biomedicines, which applies generative AI to protein design more broadly. Addgene provides a repository of plasmids including various promoters, but they are curating existing sequences rather than designing new ones. Origin appears to be unique in focusing specifically on AI-generated regulatory DNA.

The company is building what they describe as the largest proprietary dataset of synthetic regulatory sequences across diverse cell states. This is the kind of data moat that matters in biotech AI. If they can demonstrate that their synthetic sequences outperform natural ones in controlling gene expression, that dataset becomes the foundation for an entire platform business.

The diseases they are targeting, cancer and CNS disorders, are both areas where gene therapy has enormous potential but where precision is critical. In cancer, you want the therapeutic gene to activate specifically in tumor cells. In CNS disorders, you need expression in specific neuronal populations. These are exactly the use cases where better regulatory DNA could make the difference between a therapy that works and one that does not.

The Verdict

This is deep biotech, and it is the kind of company that will take years to prove out rather than months. Gene therapy development moves slowly for good reasons. But the technical thesis is sound, and the market need is clear.

At 30 days: how many research groups have started using the Switch platform? Adoption by academic labs is the first validation step, and it feeds data back into the model.

At 60 days: are any gene therapy companies in discussions to license Origin’s regulatory sequences for clinical programs? The path from research tool to clinical asset is long, but early conversations indicate whether the industry sees value.

At 90 days: what does the data look like on specificity and expression levels compared to standard promoters like CAG, CMV, or EF1a? Published comparisons would give the field confidence and attract more users to the platform.

I think Origin Bio is working on one of the most important unsolved problems in gene therapy. The switch matters more than the payload, and AI is the right tool to design switches that nature never bothered to optimize for human therapeutics.