The race begins for an FDA-approved AI toxicity model
The FDA recently announced that they’re working on phasing out animal testing requirements in favor of “AI-based computational models”, at least for toxicity screening. This is great news, both from the perspective of reducing harm to animals and lowering the cost of drug development. But this newsletter is about the life science software industry, so I want to explore another angle: This announcement just started a multi-billion dollar race to create the first FDA approved toxicity model.
This newsletter explores news and trends in life science software. If there's something you'd like to see me cover (even if it's self promotion) send me an email at scalingbiotech@substack.com. I can't promise I'll use every suggestion, but I'll try.
From a financial perspective, the best problem for a software company to solve is one where A) they directly save or make their users a lot of money and B) it’s hard for others to do the same thing (i.e. there’s a moat.)
Part B is tricky with software - The hard part is usually design and UX, but once you’ve done it, it’s relatively easy for someone else to come by and copy it. So to build a moat, most software companies rely on their brand/reputation, a well trained marketing & sales team, and a loyal customer base with a high switching cost.
Part A is similarly hard, at least for new entrants, because when there’s a big expensive problem there are usually already a bunch of solutions, written by people who noticed the problem before you did. As a problem becomes more expensive, or an expensive problem becomes more common, the software to solve it typically grows along with the problem.
Because of this, it’s rare for an expensive problem to suddenly appear out of nowhere, and even more rare for that expensive problem to allow for a moat.
But that’s exactly what the FDA did with this announcement.
Now, obviously it’s unfair to say that the FDA created a problem. It’s an opportunity by any measure. But the fact is that before the announcement, animal toxicity wasn’t an expensive problem to be solved - It was an expensive annoyance to put up with. The FDA created an opportunity to reduce the hundreds of thousands, or even millions of dollars that companies spend on pre-clinical toxicity testing. But the flip side of an opportunity is a problem.
So that takes care of part A: Animal testing is a huge expense. As of the announcement, AI models can dramatically reduce the cost. Part B comes from a mix of technical and regulatory considerations, but mostly regulatory.
As far as technical considerations, companies have been working on toxicity models for a while. In fact, shortly after the announcement, Schrödinger published a press release lauding the decision and reminding readers that they’ve been working on these models for years. Then just yesterday, Simulation Plus followed suit.
Note that these companies were building these models to solve less expensive problems like filtering out compounds that would be likely to fail toxicity testing in animal models. Those problems were indirectly expensive. Now the very same models can solve a directly expensive problem.
But regardless, it’s still software, so it could still be relatively easy for others to copy them. And we’ve recently seen that even very complex ML models like LLMs aren’t always that hard to replicate.
That means that the real moat will come from getting FDA approval. In fact, I wouldn’t be surprised if getting the FDA to approve the first drug using an AI-based toxicity model ends up costing more than building the model itself.
It’s also going to take time - potentially years. (But if it was fast, it wouldn’t be a moat.)
So it will be a while before we see who wins the race to the first FDA-approved AI toxicity model. But I promise you there are already multiple teams getting warmed up for a marathon.
Thanks for reading Scaling Biotech! When I'm not writing this newsletter, I help Life Science SaaS companies build sales and marketing systems tuned to the specific needs of a complex industry.
If you're looking for ways to connect the dots between building great tools and getting them into the hands of users, click here to request a free gap analysis.