Improving the Efficacy of Genetic Medicines

Published on: 

Shawn Davis, CEO of Liberate Bio, shares some insights into the world of drug discovery, development, and delivery, as well as process optimization

There is no shortage of bottlenecks in the drug development process, particularly for genetic medicines; to learn more about addressing these challenges and ensuring success from the business perspective, Pharmaceutical Technology® sat down with Shawn Davis, CEO of Liberate Bio.

The company leverages artificial intelligence (AI) and machine learning (ML) to accelerate drug development, as well as enhance the efficacy of these medicines through novel delivery vehicles.

PharmTech: Can you provide some background on Liberate Bio?

Davis: From the outset, Liberate was created to take on the biggest challenge in the genetic medicines space: delivery beyond the liver. When we thought about what we wanted to accomplish, we knew we had to take a new approach to the discovery process. We had all experienced the slow pace and high cost of traditional development.

For example, most molecules are tested on cells and, if they look okay and you get good engagement with the target, then you move to rodents. Then, if things look good in rodents, you move to non-human primates, which makes sense from a cost and life-saving point of view. While this approach has worked for most classes of therapeutic molecules, for nucleic acids and especially for bioaccumulation of these materials, the only way to know what works is through testing on primates.

Everything leading up to those experiments can give you confidence that your materials are not toxic and give you an early readout of expression, but to a large degree, those are well-understood issues for which experienced chemists and biologics can design from the outset. Then you get into testing on rodents, which always gets your hopes up, because everything looks like it works. You get excited, and then when it almost inevitably doesn’t translate in primates with functioning immune systems, you rinse and repeat hoping to eventually bridge the translational gap between species.

We were thinking through how to get the most relevant data as soon as possible in a cost-efficient way. That's where the RNA-barcoding platform from Michael J. Mitchell, PhD, at the University of Pennsylvania comes in; it allows us to apply the same concept of parallel processing in computing to biology.

Instead of testing one new nanoparticle design in a single animal, we can test tens or even a hundred nanoparticle designs in a single experiment in one animal because we’ve added unique barcodes inside each particle. Not only does it minimize the number of animal experiments we conduct, but we can also speed up the discovery process by testing concepts in parallel.

We no longer optimize the chemistry of the lipid, then modify the formulation to get that right, and finally add a targeting moiety if it’s necessary. We are testing all of these conditions at the same time, in the same experiment, and in a single animal, or multiple animals for the sake of proving everything works. It's very efficient, cuts the cost almost 100-fold, and dramatically speeds up the discovery process.

PharmTech: What role does augmented intelligence and automation play in this process?

Davis: Once you resolve that testing bottleneck, then you look at the rest of the design-make-test-analyze cycle. You still have to generate hypotheses (design) and synthesize materials (make) at a rate that takes advantage of your improved testing if you want to speed up the full cycle. When it comes to making materials, you need automation to be able to make hundreds of formulations in a time-efficient manner.

The 'analyze' and 'design' phases are both augmented by AI and ML. Initially, they help you synthesize the information from those experiments into actionable insights. What is it about this particular chemistry that make it work better? How do we double down on that element?

Finally, generative AI is a great tool for creating new ideas. I don't know if you've played around with ChatGPT at all, but you can input a command like 'give me 100 ideas for company slogans.' It would take me an hour or two, but ChatGPT completes the task in 30 seconds. We do the same thing, but instead of asking for slogans, we ask for new chemical structures that are similar to something we think is a good starting place.

The best description I've heard is that ChatGPT is like having an intern with infinite patience to track things down. However, this intern will occasionally lie to you. So, you still have to apply quality checking (QC). If you have infinite time, you can put anything in there. In this case, it's a bit like having a bachelor's-level chemist that would spit out thousands of ideas.

Then a PhD chemist with experience in the field looks at those thousands of ideas. They may note that a lot of it is garbage, but some of it may be interesting. AI doesn’t hand down the perfect molecule from day one; however, a chemist may go from considering it an 'atom blender' to another great chemist to bounce ideas off of. This is the 'augmented intelligence' I’m referring to. A brilliant human working with an algorithm that is still learning.

I was at Amgen setting up the first centralized innovation group for the organization with a focus on digital health, diagnostics, and drug delivery. Most people understood the latter two, but digital health was much more unknown around 2014 or 2015. A lot of people didn't understand where it was going.

Then you'd run into somebody like Philip Tagari, who was the head of translational science at the time for Amgen. His team had already figured out the five key questions they were trying to answer and were determining the questions they had enough data to work with. He was probably miles ahead of everyone else in the industry at that time.

PharmTech: In terms of how AI and ML are revolutionizing drug development, what are Liberate Bio’s next steps in terms of leveraging these tools, and are you looking to do that internally or through partnerships?


Davis: Everything I’ve described is like having tools in a toolbox that you can use to build something useful for patients. For us, we’re building delivery vehicles for genetic medicines that are extrahepatic. When you look at the golden standard of delivery vehicles today, you've got the lipid nanoparticle—billions of doses have been delivered with the COVID-19 vaccine.

Unfortunately, it's almost entirely trapped to delivery to the liver. Within the first two or three minutes of administration via IV, almost 80–90% of it is taken up by the liver. Let's say you have a magic targeting widget attached to the surface; would it even have time to do anything? I think it wouldn’t. Liberate Bio is trying to create these vehicles by first getting the chemistry right and giving it a long enough circulating half-life, so it's not taken up by the liver immediately. This allows it to redistribute to other organs of the body so that we can carry genetic medicines there.

From a business-model standpoint, we want to be a therapeutics company. The key here is recognizing the evolution of the relative contribution of delivery and the therapeutic cargo. I've been in drug delivery for 20 years, and I'm self-aware and humble enough to recognize that we're almost always second fiddle when it comes to the efficacy and safety of a therapeutic. We can make the dosing less frequent or easier for patients to administer the medication, which can offer commercial differentiation, but it’s not making or breaking the product.

That's true until you get to nucleic acids. Without an effective delivery vehicle, you don't have a therapeutic. Post COVID-19, the commoditization of these nucleic acid modalities has been stunning. I was at AstraZeneca (AZ) during COVID-19; we were working on Defense Advanced Research Projects Agency’s (DARPA) pandemic prevention programs even before that. It cost a fortune to make tiny amounts of research-grade mRNA. Post COVID-19, there can be five vendors competing for the business to figure out who can make it the cheapest, fastest, and best for you.

That really shifts the balance. If there are these commoditized mRNA—let's say you're going after Fanconi anemia in the hematopoietic stem cell—I can make an mRNA for that tomorrow with no problem. Optimization exists that could make it better, but the delivery vehicle is going to be what separates that product in terms of its efficacy. That's our basis for becoming a therapeutic company and requires us to have access to these cargos, so it limits what fields we can go into. However, it also means that we have the opportunity with those vehicles to have a second path to that business model, which is through licensing.

When you think about building a business, especially in therapeutics, you usually start with determining which target to go after and what indication it aligns with. That tells me about the patients, population, and projected value. When it comes to delivery vehicles, they're not target or indication-specific. They're cell-specific. If I solve delivery to a cardiomyocyte, I don't have one target to go after; I have half a dozen well-validated genetic mutations that I could address. Those underpin four of the largest cardiovascular indications in the world.

One of those might be for Liberate, the other three might be for others that already have therapeutic cargo waiting for a delivery vehicle. Let's say you give away mRNA for target C, and you then do a second deal for CRISPR-Cas9 for target A, and then I hold onto something else. That gives me a faster path to clinical validation, revenue, and, frankly, to curative potential for patients. So many companies have identified targets and have good cargos to go after them, but don't have the delivery. That's an important stream for us, but it is secondary to us building our own pipeline.

PharmTech: It can be difficult to place Liberate Bio on the financing spectrum. Are you a seed company? Where do you fall in that financing structure?

Davis: We are a seed company and a venture creation activity by Khosla Ventures. Nessan Bermingham, PhD, who was the founder of Intellia, is also an operating partner at Khosla Ventures. He got together with Mike Mitchell to form Liberate Bio and brought on the rest of the team. They've put in seed and convertible note funding up through now. We're doing some fundraising to do a proper seed round.

For the first 12 months, the company was establishing that platform we talked about. Then at the start of this year, we felt confident; we had done the first non-human primate study. We proved mathematically that you could test 100 nanoparticles and so we hit the ground running. We basically had a freezer full of chemistry that we had already design and manufactured, so we screened 150 novel nanoparticles in those three months. We did that not just in the non-human primates for bioaccumulation, but also in parallel with rodents for expression.

What's great about those complementary and orthogonal screening approaches is that there are many reasons they won't match up. AI suggested certain things would accumulate like crazy in a non-human primate, and they did, but then they didn't express at all in the rodent. I don't know what stopped it from working, but it didn't work. Similarly, we had some rodent data that, when I saw it, I couldn't believe. Can you get 5% only to the liver and 95% to the other organs?

When the two screens do overlap, we start getting excited. Even in the first batch of materials, that's exactly what we saw in the bone marrow, spleen, heart, and skeletal muscle. We're focused on screening broadly and unbiasedly, but you need to direct it in some way. We’re excited about cardiomyocytes, hemopoietic stem cells, and podocytes in the kidney.

This is because each of those cell types has multiple, well-validated targets unpinning indications with unmet need. They're also reasonably scientifically tractable. There is so little in the field for delivery to anything other than hepatocytes in the liver, there is value virtually everywhere else. It comes down to what you think you can do. Because of the distribution we achieve with delivery via IV, we felt confident about hemopoietic stem cells in the bone marrow. It’s a similar story for the heart, I’m personally excited about the potential for these indications.

PharmTech: It is interesting hearing about startup cell and gene therapy companies dipping their toes into the cardiomyocyte realm, but there hasn’t been much notable progress yet. Can you weigh in?

Davis: If 20 years in drug delivery has taught me anything, it’s that if you really want something somewhere, you poke a hole in it, and you put it right there. The heart is nice because, with some work, you can put it directly in the heart. AZ has demonstrated direct intracardiac administration that requires you to crack a chest open, which is not most people's preferred route of administration. But you can do intracoronary administration, go into the femoral artery, and deliver it directly to the heart, and that will avoid a lot of this first-pass uptake in the liver and allow it to concentrate in the heart.

Going back to the toolbox analogy, you have to be open-minded about what tools you have at your disposal to make sure it gets where you want it to go. Unlike bone marrow, the heart provides that opportunity. Even the kidney provides that opportunity. I didn't realize there were nine direct administration routes in the kidney.

So far, unfortunately, we haven't had much success in the kidney, so that’s a future opportunity. However, we're seeing 20–30x greater accumulation in the bone marrow than we are with the lipids published in the literature. In the heart, we see 5–6x, the spleen, maybe 5–10x. These are all really exciting, both for immune cells in the spleen as well as potential in both cardiomyocytes and hemopoietic stem cells.