Episode 151: Using AI to Bridge the Translational Gap in Biotech and Drug Development (RE-RELEASE)

On this episode of BIO from the Bayou, we're not monkeying around. Learn how to connect those monkeys to humans using AI. You're listening to BIO from the Bayou, empowering biotech leaders with the tools, assets, and expertise to improve healthcare outcomes.

Each episode provides insights and information to expand your network, advance your product line, and help your company thrive.

Serving the entire Gulf South, but recorded in the heart of New Orleans. Welcome back to BIO from the Bayou. I'm your host, Dr. Elaine Hamm, the Executive in Residence at Tulane School of Medicine, and we We are actually Bio on the Bayou this week. We are taking advantage of our conference this week and having an opportunity to sit down in person with our incredible guests. And our guest today is Dr. Todd Kilbaugh. He's a professor of anesthesiology, critical care medicine, and pediatrics at the Children's Hospital in Philadelphia. But for Bio on the Bayou, he is coming actually as the founder of Pharos BioLabs. And I'm excited to learn— well, I already know a lot about Pharos BioLabs, but I'm excited to share what you guys are working on with our many, many guests. Welcome to the show, Todd.

Thank you so much. It's a pleasure to be here, and I'm excited to be on your program.

Okay, so what is an anesthesiologist, critical care doctor doing with a startup called Pharos Biolabs? Tell us the inception or the origin story.

Yeah, the insidious story.

Where were you? What were you drinking?

My career really started here in Louisiana. So I was a pediatric resident here at LSU Health Sciences. for several years. So, I started all my training here, fell in love with the city.

Yeah.

My life changed over time, and I ended up in Philadelphia, where I finished my training at the University of Pennsylvania and then Children's Hospital of Philadelphia. And during that period of time, I started working in biomedical engineering and just sort of fell in love with the idea of science and research.

Yeah.

And one of the things that I really noticed was that one of the major problems in pediatrics, in fact, the number one cause of morbidity and mortality in pediatrics is traumatic brain injury.

Really?

No FDA-approved drugs for traumatic brain injury. Bigger problem than pediatric cancer, bigger problem than fresh water, bigger problem than infectious disease.

I had no idea. I mean, when we think of TBI, we always think of adults, but yeah, children.

So in a very rapidly urbanizing world where in the United States TBIs have gone down, the rest of the world traumatic brain injury has gone up. So huge problem for our warfighters. As you know, it's a signature injury for most of the last several conflicts that we've had, and also still a serious problem for children. So I got really interested in sort of the gap in translation. Why was there zero FDA-approved drugs? I also fell in love with mitochondria at the time.

Sure, that happens.

It happens, you know, you're sitting around in a bar.

You meet a mitochondria in a bar.

That's right.

There's a joke in there somewhere.

Yeah, exactly. 3 mitochondria walk into a bar, what happens?

Who's your mother?

Exactly. Oh, good one.

Yeah, nice science.

I'm very lucky that I started working in an institute with a gentleman named Doug Wallace, who's won the Lasker Award for mitochondrial biology, really is an incredible mitochondrial biologist, and sort of sparked my love over the next sort of 15, 17 years of developing therapeutics, devices, and diagnostics in sort of this translational gap, which I really felt was a problem in making the jump from rodents to humans.

Yeah.

animals are not the same.

Turns out mice— I think the phrase is mice lie, monkeys exaggerate, and humans just sometimes fail. So making that translational leap is so hard these days. I mean, throughout the history of the development of drugs and therapeutics, that next step is so hard.

Yeah, most things fail, right? So if you look at sort of even funded companies, about 80% of the ideas for admet for toxicology, distribution, metabolism. Most of that doesn't translate from rodents into non-human primates. And so I really focused on porcine models for a long period of time because of trauma, and that sort of built out to extracorporeal support devices.

Yeah.

Cardiac arrest, traumatic brain injury.

Yeah.

All kinds of things in my lab just kind of grew and grew and grew.

Yeah.

And sort of the inception of Pharos was really, how do I kind of have a bigger ripple effect outside of even my own lab? And that's how I started looking for collaborations and those sort of things of how we could actually really change the translational gap, the whole translational idea of taking an idea from conception, and maybe you are using rodents or humanized mouse models or even mathematical models in silico.

Mm-hmm.

to come up with ideas of what might work and what might not work in terms of drug distribution and drug efficacy?

Yeah.

And how could we then, in this new idea of AI organoid development, reduce and reuse and do efficient non-human primate studies?

Yeah.

How could we really just sort of really set the whole preclinical development pathway on its ear?

Yeah. Well, and we're really being pushed by this, especially this administration, of moving away from animal models. I mean, I've always wanted to move away from animal models, but we are— I don't know if— I feel like we're not quite there yet. And we have, of course, a lot of things in the AI world that are helping us get there. Do you think that— how far away out do you think that we are from getting rid of animal models completely?

A long time.

Yeah, I do too. I wish it were different, but I mean, we want to do it right.

Well, I think that's the problem, you know, like, AI models in general, they're taking data that already exists and searching faster and giving you answers faster. But the problem is the data doesn't exist.

Doesn't, yeah.

So an idea for lab-on-a-chip, organoid models, those sort of things, all fantastic, all part of an FDA modernization, and this is the goal of what we want. The problem is, how do you do that safely? How do you maintain regulatory optimization? How do you, in this next phase, really, how do you use animals appropriately optimize their lives, not only for human health, but I think we also have to realize that most of the drugs that are available for veterinarians, for others, come through what we develop with animal models.

Absolutely.

So how we deal with our farmers, how we deal with production animals, how we deal with our companion animals, all of these things come through animal testing. which we want to reduce. And the ability to do that has to start with building the right organoid models, building the right in silico sort of ideas, building the right mathematical models that then take over, over time, the use of these animal models.

Yeah, yeah.

So it's great to say it, and it's the wish that we want, but it has to be done in a way that preserves the science for everyone. And I think that's really the key, and that's what we want to focus on at Pharos.

Yeah, and that, you know, we talk about failure. I mean, efficacy, you could have— I mean, I've cured a lot of mice of MS.

Yeah.

I've cured mice of endometriosis. Mice don't even have periods. So where do you think that we fail? This preclinical stage that I operate in, that you guys operate in, We have, what, the 10% chance of actually becoming a drug, if that? What do you see beyond just efficacy— like, I have efficacy— that causes a failure from now until we get it into a patient? And it can be anything. It doesn't have to be just the science part either.

Yeah, there's so much. Study design is one of them. Let's take traumatic brain injury, for example.

Tough indication.

So a tough indication, very difficult to get venture capital interested in funding it. You say it out loud, it's a huge problem, right?

It is, huge market.

You can say the total addressable market is X, it's the number one cause of disease and disability in our warfighters. Like, you can just say all these things.

Market's there.

Everybody goes, yes, yes, yes. But try to move an asset forward. Yeah, not interested. Yeah, it's mostly because you can't do good clinical trials, right? So hard. So where you have the oncology field, you know, your oncologist, you don't walk in and they say, oh, by the way, You have breast cancer. They say, you have breast cancer, we've done a biopsy, we know the genomics of this, we know the trials we can put you in, what you won't be put in, and then we're going to follow biomarkers as we go that will tell us whether we would modulate you on one oncology pathway versus another versus another versus another.

Endpoints are much more clear.

Cancer go away. Endpoints are clear and biomarkers are clear. on when to change and when to adapt. And so adaptive trial design is a huge problem. And I think this starts in our preclinical phase when we don't use things like Olink proteomics and we don't use— and we're not developing biomarkers as we're developing indication to follow these particular assets forward into clinical trials.

Yeah.

Those companion diagnostics teach us a lot about whether they who to enroll, when to enroll, the differences between genders, the differences between age, all of these different things that we sort of, we guess at in a lot of, especially the CNS world.

For sure.

Has been a huge problem. And I think that starts at the preclinical phase. And again, using AI models to understand what the biomarkers are in humans, how that looks in primates, how that looks in porcine models, how that all sort of works together to actually predict your clinical trial so that we have a fit instead of a 15-year phase of drug development, we shrink that to 5.

Yeah.

And we do it really quick, tight, and it's telling us what to do in our clinical trial.

Yeah.

The cancer field has that on lockdown.

Yep, they do.

How do we take that to everybody else? And they've done it really well, and I think the gene therapy groups and those sort of things for rare disease are starting to figure some of these things out too. And that's gonna be key to what we do in our next phase of clinical development.

Because oncology has the great luxury of having gobs of data out there, and years of it being really focused on gathering data. CNS, probably less, I mean, definitely less so. I would say probably more quality of the data, especially as you get into some of the more genetic-linked sort of rare disease CNS types of things. But I mean, I can't tell you how many times I've had investors say, what biomarkers are you gonna be looking at in the clinic? You better know these days your biomarkers, and you better be looking at it in mice and in monkeys, and see what they can do, and follow it throughout the trials. And so having that, knowing your prognostic or your diagnostic that you're going to be utilizing, you need to know that in your preclinical stage, definitely. And so, and I think that's where really the AI piece comes into play in helping us bridge that gap.

I think so too. And that mathematical modeling, again, venture capitalists are smart.

Yep, very.

They understand how to model. They understand how to write Python. They understand how to do financial modeling. They are very, very, very, very, very good combatants. And so when you as a scientist sort of go in and you forget that their math is as good as your math—

Yeah, it turns out.

It turns out, if not better, being able to break down mathematically why your asset will work and why it will not just work from an efficacy standpoint, but why it will pass a phase 1, How it will work in a Phase 2a/2b, and how then the mathematical modeling of how it will work to a Phase 3. And that starts, I think, in your pre-IND. And that's where I think the CRO industry fails us a lot. The big CRO industry is really built for you to go as a small, medium-sized company to go back a few times because the studies didn't quite work. Quite work, yeah. You really need to do it again. It turns out every single time even you fail a toxicology study, that's a 6-month revamp if you're lucky.

If you're lucky.

It's probably $1 million or what people suggest is about $1 million of burn rate for every 1 to 2 months that you failed your first initial toxicology study. And that's just tox. We're not even talking about efficacy.

Yeah.

So, this big valley of death, there's an engine that's sort of there for you to fail. It's made for you to fail because there's about 100 consulting companies that are living off of you. There's the CROs that are living off of you, and they want you to keep needing them and those sort of things. And I think that's where this AI revolution sort of comes in, where—

Definitely.

If smart people, including venture capitalists and scientists and entrepreneurs, can capture good data quickly, understand that modeling, and do it in a, the right way, then everybody's talking on the same plane.

Yeah.

And I think, quite honestly, a lot of those consulting companies and a lot of those others start to drift away. And I think the trick in this is having AI co-scientists and AI co-regulatory. You need the really smart scientists, and you need the really smart regulatory people, and you need the really smart business people.

Yeah.

that work with your agentic AI, that just gives them data that they can interpret faster, quicker, cleaner, and better. And that sort of merger of those things hasn't quite happened yet, but it will. And the companies that do that, whether you're a venture capital company, or you're a scientist, or you are—

A CRO.

A CRO, it won't matter. It matters if you do it right and do it fast.

Yeah, and that piece around— so as we think about, it's just nice to have that in your back pocket when you have an investor that asks not just the biomarker question. If you've gone through something like a digital twin process where you can try to predict side effects and say, look, we're about to make that big leap that all preclinical companies have to make in going from a monkey into a human, but we have this extra set of data that we've done, and it's early days, but we think that the side effects that we're gonna see is this. If you're lucky, you can maybe throw in a competitor's drug as well and do the same kind of comparison. comparison. And it just adds to that confidence of the story as we sort of move into a new type of science where we're able to take those AI tools, pair that with the work of a CRO, and utilize those monkeys appropriately and be able to help patients faster, and hopefully prevent some of the side effects and things that kill drugs in the long run.

For us, that's where I'd like to see Pharos go. The idea of using—

Yeah. Tell us where you guys are headed.

Well, I'd like to see AI, organoids, and non-human primates, sort of in that sort of closed-loop validation, over time reduce, reuse, and primates disappear.

Absolutely.

As part of our ecosystem for preclinical development.

Yep.

That will happen if you develop the right organoid models, the right in silico models that are based off of, for example, a liver. We want a liver on a chip. That sounds fantastic.

Sounds great, right?

Amazing. And the FDA says you can just use that and take that directly into a clinical trial. I'm not sure as a patient I want that.

I don't want to.

I'm not sure a venture capital group is really ready to bet their house on a liver on a chip.

And certainly not pharma. At the end of the day, pharma is still going to want all of that.

Of course. And so What we want to do is create the correlation and the mathematical modeling between the best organoids and non-human primates that tell us that this is safe.

Yeah.

And then use the least amount of non-human primates to then do the best studies that predict in humans, not just the toxicology, but also the efficacy.

Yeah.

And I think what'll end up happening over time is we'll use less and less non-human primates for toxicology, and we'll use more non-human primates for proprietary disease like Alzheimer's.

Yeah, more the niche type of things that are terribly complicated.

Terribly complicated. And so your interest is obviously in women's health, and we've spent no money and no time dealing with how that even works at all.

Yeah, we've been studying everything in male mice and males and blood.

No money.

Yeah, yeah.

Outside of rare disease for pediatrics, what have we done in kids?

Almost nothing. Kids are scary for drug developers. Of course they are. Yeah, yeah, and so are women, 'cause they can get pregnant and it might hurt the baby.

There's so many, so many variables. All of those things. And I think that's, we haven't even started to build the datasets for the generative AI to go out and, collate and feed through and do that. And so until that's actually built, then we have— we won't really have anything.

But it feels like we're on the cusp of that. And I really am thrilled of groups like Pharos of being able to take all of these things, these resources that are out there, and not just say slap AI on your pitch deck and say, yeah, we got AI, but really thoughtfully using that in parallel to the resources that are also cutting edge, like organoids, and being able to piece these things together to a quilt that actually can be useful for moving things that really big step that most companies are at right now. So that's exciting stuff, Todd.

Yeah, I think I'm very lucky that I've been a physician and a scientist for a very long period of time, and it's a really exciting period, I think, over the next couple decades. Yeah, we've watched immunotherapies just kind of explode oncology. Yeah, and for me, I've watched the— especially around the CNS space just kind of eke along and eke along and eke along. And I think with these types of changes, our understanding and developing organoid blood-brain barriers and AI to predict that and using non-human primates effectively and thoughtfully and those sort of things for a very complicated deep blue ocean of the brain.

Yeah.

We're gonna see that inflection point over the next decade to 2 decades, especially with biomarker development, all these things. The ability to do these types of real adaptive trial design, quickly take drugs, you know, Alzheimer's drugs right now, the ones that are getting funded, they changed outcomes by 2 to 3% and they're billion-dollar drugs.

And can make it worse if you are ApoE4 genotype and you have that indication. It can be hard.

And that's how starved we are. We'll take that, how desperate we are to solve The brain.

Yeah, I'll take a DMO because I know.

100%.

Yeah.

I think it's just a really exciting time, and I'm happy to be a part of it.

Well, I'm thrilled for your brain, someone that can help that, you know, again, knit that quilt together of these wonderful resources that are out there and use them in a really thoughtful way to make those big leaps, and it is a really exciting time and an exciting time to have a group like Pharos to help lead that charge, so thanks for being on the show, Todd.

Thanks for having me.

Great.

Appreciate it.

As always, we'll have more information about Todd, about TBI, about Pharaoh's Bios in our show notes. Check them out there and be sure to like and subscribe and learn more about Bio from the Bayou. Thanks, Todd. Thank you.

Thanks for joining us for Bio from the Bayou, and we hope you'll join us again.

To never miss the latest updates in biotech, hit the subscribe button now.

We'll catch you on our next episode of Bio from the Bayou.