Dr. David Fajgenbaum

And early on, as you'll remember, there was a lot of drugs that were repurposed.

Some worked, some didn't work, but there was a lot of repurposing.

This is the first time we did like a very concerted effort to be like, what else is out there for this one disease?

Very much informed by what we'd done previously.

And COVID, of course, there's lots of controversy about what worked and what didn't, but the two drugs that unquestionably worked incredibly well were dexamethasone and tocilizumab.

They saved millions of lives and they were, you know, old drugs have been around for a long time.

And so that further got my wheels turning on like, what if we could create a system to automate what my little lab was doing for one disease, but we did it for all diseases and all drugs simultaneously.

And thankfully, in parallel to those dreams, the field of machine learning, artificial intelligence has matured so much that we can actually do that.

So in my case, you know, you can think about this.

We use what are called biomedical knowledge graphs, which are just sort of mapping out like every medical concept on a map.

So you can imagine like if you have this giant wall and

and every single gene, every disease, every protein, every pathway was put against the wall.

So if we were to start with that concept and say, well, what do we do for me?

Well, you'd find Castleman's on that wall.

It would only be there in one place.

You'd find Castleman's.

And what you'd find is you'd find an edge or a line between Castleman's and activated T cells, because I discovered that T cells were activated in my disease.

You'd find another line to mTOR activation, because I discovered that mTOR activation was really up in my particular immune cells.

And then you would find a drug from T-cell activation and mTOR activation to serolimus.

Because serolimus is able to inhibit mTOR activations and able to inhibit these activated T-cells.