Generate Biomedicines, Inc. Common Stock H.C. Wainwright 28th Annual Global Investment Conference

Thanks, Mitchell. Great to be here.

Sure. Generate Biomedicines was founded eight years ago, before generative technology, generative capabilities were really even talked about, and well before ChatGPT and AlphaFold and some of those other technologies. Really what we were founded to do is use generative machine learning technology to ultimately discover innovative, differentiated, novel protein therapeutics for patients. Really trying to discover things that traditional drug discovery is unable to do, just given the ability of technologies to explore a much broader landscape of potential opportunities for proteins. What we did very early on at the company is started to understand what the technology could do, and one of the big concerns, or at least ideas that could have been concerning is are these novel therapeutics that are generated by AI or machine learning, are they immunogenic?

How would they actually be handled in the body relative because they're not formulated using the traditional technology or traditional techniques. We started very early with looking at low biology risk products, which is why TSLP, our lead asset, was even just talked about and we initiated on. Low biology risk, known pathways, safe molecule, but had some deficiencies in it that could be improved upon that could improve a patient's experience with a medicine that was very highly effective and safe. Tezepelumab was the first, or anti-TSLP molecule, was the first lens that we took as a company. What we did with that molecule is effectively used half-life extension technology, which a number of other companies are certainly using, but used our technology to change the molecule to get a much better binding affinity to the cytokine.

While we hit the same epitope as tezepelumab, we have tighter binding affinity by 20-fold, frankly, 100 femtomolar level of binding affinity. We think the critical nature of that molecule is not just the long half-life, which is a 98-day half-life, but also you need that tight binding affinity on the cytokine to ultimately get six-month dosing and not have waning efficacy over time. That's why we started with TSLP. Ultimately, we've done multiple partnerships with Amgen and Novartis as examples, who we would view as some of the greatest protein engineering companies out there, and they obviously have been extraordinarily successful at using traditional techniques and coming up with molecules.

What they did in our partnerships, Amgen we partnered with five years ago, Novartis we partnered with two years ago, is they're bespoke around individual targets where ultimately they gave us some of their most challenging, or most difficult challenges with protein engineering to come up with molecules or test biology. These are target-specific partnerships, but what it taught us, in addition to the great partners that we have, it taught us where the struggles that traditional techniques have in protein drug design. I'll get into some of that because it really gets to the platform and the next wave of products that we have. Ultimately, we've learned from that, and some of the new wave of things that we're working on are those innovative technologies.

One of the core things that differentiates us, and I'll come back to the platform, is we're not just a machine learning AI company that uses models and ultimately believes we're going to come up with the next wave of therapeutics based on the quality of our model. We think models are ultimately going to commoditize, are already commoditizing. Frankly, just having a better model, and even though we think we were first and probably have one of the best models out there, that's not enough. Biology is just such an unknown science, an unknown space. If you think about large language models and using the internet, if you ask ChatGPT to write you a paper, it can do an incredibly great job at writing you as long a paper as you want on almost any topic. Biology is just not known.

It doesn't have the breadth of knowledge. It's probably only 5% or 10%, maybe 15% known. It's not like you can use LLMs to come up with billions of hypotheses and know which one is actually going to be the right hypothesis. You need something to be able to test, an experimental wet lab capability to be able to test at scale these molecules. One of the core differentiators for Generate is while we've invested heavily on the models, we've invested as much, if not more, on the experimental side. While we can get billions of hypotheses, we actually, in the order of days or weeks, can test every single one of those hypotheses in bespoke assays to determine which are the ones that are going to be most promising for protein therapeutics. Happy to get more into that.

The last point I'll just mention, because there's a lot more, we've translated five molecules from our model into the clinic. We've shown they're not immunogenic. There's no more ADA formation, actually very low ADA formation. They're highly efficacious molecules. We have our lead assets are in phase III, which is the anti-TSLP. We have a couple oncology assets that are in phase I, which happy to talk about as well. That next generation of molecules and where we're focused now are on the things that China can't necessarily brute force their way into today, or traditional techniques really struggle to do. These are things like pH-dependent binding. Can you get 1,000-fold difference in a protein binder at a pH of 7.4 versus a pH of 6? Internalization. We're working a lot with ADCs or other molecules where you have great targets, but very poor internalizers to be able to kill cancer cells, as an example.

We've developed techniques that can get logarithmic fold improvement on internalization of these antibodies or other protein crossing barriers, blood-brain barriers, cellular barriers, gut barrier as another example. Using these types of techniques, we have a number of programs that are on the come, other than just the ones in clinical development, where we think we're more highly differentiated.

Sure. TSLP is a known mechanism in asthma. TEZSPIRE has been on the market for five years. It is a biologic. Biologics in respiratory and asthma particularly are very low penetrated, so it is about 20% penetration in asthma. The great thing about the TSLP molecule, as has been shown through TEZSPIRE, is it is an all-comers of eos population in the asthma population. What you saw with the TEZSPIRE data is a 70% reduction in exacerbations in the high eos population above 300 eos, and about a 40% reduction in the low eos population below 300. Overall, the population, you saw a reduction of 56%. It is a monthly dosed medication in asthma, and so the benefit for our drug over that is it is a six monthly dosed drug. We hit the same epitope, as I mentioned before.

We have chosen our dose and moved from phase I to phase III, which I will talk about, and why we think we are very confident with that. In 300 milligrams, we are seeing 99.9% saturation of the target, which is the cytokine in this case, which is about the same as the approved dose in tezepelumab, the 210 milligrams also has 99% saturation. So there is no reason to believe there will be better efficacy in the clinical trial. We did show preclinically a five times better potency, but again, if you are 99% saturating a target, that you are hitting the same target, the likelihood of seeing better efficacy is probably small. In the real world, however, what has been shown, and some data were presented at ERS last year, is that only about 20%, or it is suggested that only about 20% of patients on biologics in respiratory are adherent to their medicines.