SPEAKER_04

The body has evolved a mechanism to create what essentially looks like a lymphoid structure in the middle of the tumor.

It's sort of a forward camp of immune cells that the more of those you see in a tumor, the better will be your outcome as a patient.

And so we used a cohort of colorectal cell, basically colon cancer patients, where we looked at hundreds of biopsies.

And we did that pseudotime analysis where we looked at

for mature tertiary lymphoid structures.

And then we looked for immature, slightly less mature, even more less mature, etc.

And we were able to backtrack to the cell types which need to come together that would then form the more mature.

What use is that?

It's a nice paper, but it also now tells us what we might do to create more of these in a tumor.

Because we already know from multiple kinds of tumor types now that the more of these tertiary lymphoid structures you have, the better off will be your outcome with chemotherapy.

So it might be, for instance, that once we know that you have a disease like this, we could give you some kind of therapy, a virus or what have you.

that goes and homes to the tumor, seeds the beginnings of these initiators with, there's these cytokines that are produced that are necessary for initiating the formation of these objects.

And so there's a huge benefit to that, but we never would have found those, in my lab at least, without the AI.

Because it basically did the work for us.

Correct.

So we use, well, we can use pretty much any of the LLMs, but right now we find that OpenAI is the best.

For us, at least.

And then we create an agentic overlay.

Basically, what's called, you probably know, chain of thought, which is a series of questions.

So how we taught it was we basically came up, here's a hundred kinds of questions a scientist would ask about the immune system.