We have the tools to edit genes and design drugs, yet thousands of rare diseases remain untreated. The problem is not a lack of science, but a lack of scientists. There are simply not enough human hours available to test every possibility for every condition. Two biotech companies are now arguing that AI is the only way to clear this backlog, not by replacing biologists, but by doing the brute-force calculation that humans cannot finish in a single lifetime.
Key Takeaways
- GenEditBio received FDA approval to begin CRISPR therapy trials for corneal dystrophy.
- Insilico Medicine launched the MMAI Gym to train generalist models for drug discovery tasks.
- The FDA approves approximately 50 new drugs on an annual basis.
Insilico Medicine and GenEditBio are using machine learning to automate the most tedious parts of biology. Insilico is building what it calls “pharmaceutical superintelligence”—systems designed to do the work of legions of chemists. They recently launched a training ground called “MMAI Gym” to teach generalist AI models how to handle complex drug discovery tasks.
GenEditBio is focusing on a different bottleneck: the delivery mechanism. They use AI to predict how to get gene-editing tools into specific tissues without the body fighting back. Both companies are betting that software can identify patterns in biological data that are invisible to the human eye.
The big deal
The pharmaceutical industry currently operates at a crawl. The FDA approves only about 50 new drugs per year. That pace is too slow to address the thousands of rare disorders that currently have no treatment options. Because developing a new drug is expensive and labor-intensive, companies usually focus on common diseases that promise a high return on investment.
If AI can automate the early stages of discovery, the cost of finding a new drug drops. This matters because it changes the economics of medicine. If it becomes cheaper to design a drug, it becomes feasible to cure rare diseases that big pharma usually ignores. We are also seeing a shift toward “in vivo” treatments—medicines injected directly into the body—rather than the complex process of removing cells, treating them in a lab, and putting them back.
How it works
GenEditBio uses AI to solve the “delivery problem” of gene editing. The medicine needs a vehicle to travel through the body to the right cells.
Think of gene therapy like shipping a crystal vase to a friend. The vase (the medicine) might be perfect, but if the delivery truck crashes or the package is smashed on the porch, the vase is useless.
In this scenario, the “truck” is a nanoparticle that carries the gene-editing tool. GenEditBio uses AI to simulate thousands of different chemical trucks to find the one that can drive straight to a specific organ—like the eye or liver—without the body’s immune system attacking it. They test these predictions in a wet lab, and the results are fed back into the system to teach the AI how to design better vehicles next time.
The catch
These systems are only as good as the data they learn from, and biology has a data problem. Executives from Insilico note that current medical datasets are heavily biased toward Western populations. If an AI is trained primarily on data from one group of people, the drugs it designs may not work effectively for everyone else.
There is also a scarcity of “ground truth” data. To model the edge cases of human biology, researchers need massive amounts of high-quality patient data, which is difficult to obtain. While the companies aim to build “digital twins” to run virtual clinical trials, they admit this technology is still in its infancy. You cannot accurately simulate a human body if you do not fully understand the instructions that built it.
What now?
GenEditBio has received FDA approval to start trials for a treatment for corneal dystrophy. This will be a major test of their AI-designed delivery vehicles in real patients. If you follow the sector, watch to see if Insilico’s generalist models can actually outperform specialized tools in their new “gym.” The industry is watching to see if these digital simulations translate into safe, physical cures.
