11.8.22-Tokyo-Showcase-Aria-Pharmaceuticals

ANDREW A. RADIN: All right, and just to put our work in a little context for you, to create new drugs in the marketplace, you need exceptional chemistry which is what our previous speaker Roy just told you about. But you also need a deep understanding of the biology of diseases. And that's where we work at Aria Pharmaceuticals.

Company started with my studies at Stanford University. We also have co-founders and a founding team from MIT as well. Now, biology, and especially very complex disease biology, is difficult to understand. And data is how we in the modern age are able to better understand these very difficult disease pathologies. And analyzing this data is critical to be able to create new insights in each of these diseases.

But it's very difficult to do, because the data sources we have today, there is no single data source in biology that explains all of disease pathology. And the data sources that we do have, each of them represent a very narrow window of information about disease, either directly as a biomedical measurement, or indirectly as a phenotypical measurement. And analyzing all of these data sources together is critical to better understand these very complex diseases.

Now, what we're able to do is combine these orthogonal and heterogeneous data sources and detect signals that other people can't see. Now what most people do in disease pathology is they look at a single puzzle piece, if you will. They collect data that they believe is going to unlock some biological mystery in a disease. And then they process that data looking for signal.

Now, these approaches most certainly do work. But they only work if the answer to your biological mystery lives within that data set, OK? So here on the lower side you can see traditional approaches. And you can imagine this is like a real physical puzzle that you have in front of you. If you only had a single puzzle piece and that puzzle piece was a very crisp picture of the face of a rabbit, you don't necessarily need the rest of the puzzle pieces to understand the creature that you're dealing with, all right?

But you could imagine you collect that single puzzle piece and it's a picture of the sky, or a picture of a field. OK, that doesn't tell you enough about what's going on in that disease pathology. And so what we're able to do by collecting all of these heterogeneous and orthogonal data sources is do two important things. The first is to identify the puzzle pieces themselves, and then ultimately assemble them, OK?

And when that assembly is complete you don't necessarily need a very crisp view in any one puzzle piece. And you could imagine, for example, in that puzzle you see the outline of a rabbit, OK? And that gives you enough information to better understand that disease pathology and how you could potentially address that disease.

Now, we're able to do this because we've solved a computer science problem. Most people are using artificial intelligence, of course, to examine these very noisy data sets. But artificial intelligence, as it's been designed out of the box, assumes you've got a single input type. It's designed to work with homogeneous data.

And the problem that we've solved is allowing heterogeneous data to be injected directly into the insides of deep learning models, directly into the hidden nodes themselves. And the reason that's important is because now we can process all of this data as a single unit, right? Other people might claim that they're using a diversity of data but really what they're doing is the top thing over and over again and looking for overlapping evidence.

It's only when they're combined in a single computational unit that we can uncover things that other people can't see. We've use this technology across a broad range of disease areas. Our most advanced programs, in lupus, idiopathic pulmonary fibrosis, and chronic kidney disease, we're planning to bring to IND enabling studies later this year. We have ongoing collaborations here in Japan with both Ono and with Santen.

I want to show you a quick example of our work. If you are a pharmaceutical executive, this is a very exciting slide. If you're not in the pharma industry it's very confusing. But I'll explain it to you quickly.

So what we're looking at is some preclinical results in our work in lupus. And if you're not familiar with lupus, lupus is a disease where your immune system attacks your organs. So on the mild end, inflammation, that sort of thing, in the more severe cases, ultimately you get organ damage.

And so here we're looking at kidneys. And just quickly across the different colors here, so the gray is a vehicle. It's a placebo. The animals have been treated with something that has no therapeutic effect. The red one here is cyclophosphamide. So cyclophosphamide is a drug you give in very severe cases to lupus patients.

It's actually quite effective but unfortunately it has very terrible toxic side effects. So you can't take it very long. The next two things on there are experimental molecules 711 and 712. And what's super-interesting about this is that we get similar efficacy to cyclophosphamide, especially with 711, but without the toxic side effects, OK?

And therefore we have the potential for this disease-modifying experimental medication to become the new standard of care for lupus patients. Now we've had similar results like this in other disease areas as well. And certainly, if you go to our website, you can see more scientific material about all of our disease programs underway.

So we're here in Japan specifically to talk about licensing. We're very interested in outlicensing these programs with regional rights in Japan, Japan, Korea, and China, Asia, something like that. And if you're in the pharma industry and interested in talking about in-licensing assets at pre-IND phases, I'd love to chat with you. Thanks for your time.