Good morning, everyone. Welcome to our day one of the Cantor Fitzgerald Global Healthcare Conference. My name is Li Watsek , a biotech analyst here at Cantor Fitzgerald, and it's my great pleasure to introduce our next presenting company, Acrivon Therapeutics, and we're pleased to have the company's CEO, Peter Blume-Jensen, joining us. Peter, before we get into the pipeline, I would love to have you introduce the Acrivon story, especially your unique AP3 platform, which I think is quite special. Its phosphoproteomics-based precision medicine platform, and maybe tell us a little bit more, what have you demonstrated so far to support this platform can do what you say it would do?
Thank you so much, Li, and thanks for having us here today. I'd be happy to. Everything at Acrivon is built around the AP3 platform. It stands for Acrivon Predictive Precision Proteomics, and that's what it is. It's AI-driven, and that's driving our entire pipeline that's very rapidly expanding. We have demonstrated and proven it in all the areas we set out to do. We use it for indication finding, so we identify sensitive indications before we actually enter the clinic. We've done that for all three assets so far. One of them is not in the clinic yet. We have two clinical stage assets in phase II. We have also shown that we could identify patient responders individually in a prospective manner. That's in our registrational intent trial with the ACR-368, the CHK1/2 inhibitor from Eli Lilly.
We have also published a lot on uncovering resistance mechanisms and so forth, which enable us also to rationally design drugs for superior single-agent activity, and that's what we did with our first-in-class dual WEE1/PKMYT1 inhibitor, which was designed to also activate PLK1, so we get a massive proapoptotic cell death. The platform itself is really able to uncover unbiased at high resolution at the protein level. We use phosphoproteomics, which is a way to measure activity states. We use it to look at what a drug does inside a cell. So we very quickly pass that cellular boundary, so it makes the path to the clinic very linear. There are very few surprises, and it can be applied at all stages of drug development. We've used it for rational drug design, as I said.
We can take early hits or leads, put them into cells, and we can see what they do at the signaling pathways and relate it to the disease-driving mechanisms. That's why it opens up for predicted biomarkers, massive amount of pharmacodynamic biomarkers, et cetera. What is also unique is that it's a massive amount of proprietary data we have generated for over 100 compounds now, DDR inhibitors, in-licensing assets we are assessing and comparing with competitor assets that are marketed or in clinical development. Also in the I&I area, we've done that. We did it for the Lilly asset we in-licensed. The platform really has a number of applications, and they are all proven, which is most important. So it's really, we have proven everything we set out to do with it.
The final thing I want to say is that it's, in a way, I would argue it's a first-in-class platform in the sense that we are not aware of anyone else taking such a systems approach where you can look at the whole signaling network inside a cell and just unbiased see what a drug does, what resistance mechanisms kick in, how it inhibits, and how potently the disease-driving mechanisms, and just unbiased match the mechanism of any drug with these disease-driving mechanisms. Our resolution is extremely high, close to 200,000 quantified map phosphorylation sites. We run quadruplicate samples, and we get these actionable data outputs. We have all kinds of interactomes the scientists can access, and we can make sense out of this. That's where the indication finding also comes in. Probably said more than I should, but that's really driving our platform and pipeline.
Sorry, our pipeline. We are in phase II-B registrational intent with our in-licensed CHK1/2 inhibitor from Eli Lilly. We are in phase II expansion phase with our first internally developed first-in-class dual WEE1/PKMYT1 inhibitor, and we are now aiming for an IND early next year with also a first-in-class CDK11 inhibitor, where we have identified aggressive AML as a very sensitive indication and proven that pre-clinically. I think that covers it without going into technical details.
Yeah.
But it's really based on how the pathways are affected inside a cell. So when we enter the clinic, we can see whether the drug, which was designed to do certain things at the pathways, actually do that at the first dose level, and that we also did with our first internal developed asset. We have a clinical PD AP3 assay as well to support our phase I trials. So lots of applications.
Yeah. So it seems like you guys have a very unique AP3 platform, first in class, and then you also proven yourself that you can generate a lot of clinical assets, and you have some pretty nice clinical validation already. I wanted to follow up on AI. Obviously, this is a buzzword in-
Yeah
drug development right now, and we've talked extensively about how AP3 utilizes AI, just given there's a lot of data points, right?
Yeah, sorry.
I just wonder if you can talk a little bit about your approach to AI, how you apply it, and how are you guys using it so differently than everyone else
Yeah
in the industry.
Most healthcare biotech AI companies, they focus on how to, in a streamlined manner, it is oversimplified, obviously, what I am saying, but how to generate the best possible drug in the most efficient manner against a particular target. We take the opposite approach. We want to develop or match any drug with what it does inside the cell unbiased. If it is at the discovery phase, we can then optimize the drug based on the pathway effects. If it is an in-licensed asset, we can make sure we bring it to the patients that benefit from that drug. The AI is used uniquely in the sense that it is proprietary data that I had mentioned before. All our learning is on our own proprietary data, and everything is drug-centric. We obviously triage our AI capabilities and internal proprietary, massive amount of data. We are talking about terabytes of terabytes of terabytes.
We triage that against publicly available data as well. But the AI learning environment is actually based on our proprietary data, which is very, very unique, entirely, and not relying on external data. They do not lie. We compare, obviously, benchmark against clinical-stage WEE1 inhibitors and PKMYT1 inhibitors and CHK1/2 inhibitors when we in-license that, or INI1 inhibitors in the areas we are building up now. We just see directly which drug is best, has the most desirable pathway effects, and then we can both develop drugs based on that and also choose the ones for in-licensing that are best. The whole platform purpose is to overcome the limitations of genetics-based precision medicine. The reason we in-licensed the Lilly asset and had a chance for that, which was a flagship program, was they used state-of-the-art genetics genomics and concluded that they could not identify patient responders.
Genetics did not work. We proved that we could identify. We have had three prospective data cuts. Unfortunately, it slowed down enrollment, and that is what we paid a big price for. It requires a pre-treatment tumor biopsy, and that leads to how we got into serous endometrial cancer.
That was a very good use case for the AP3 platform. I know oncology is what you are currently focused on, but you also mentioned you want to move into I&I.
That's correct.
in the future. Tell us why.
Very near future.
In the very near future. Tell us why that's a good idea.
Yeah
and maybe share a little bit about the work that you've done so far.
The reason is really where I ended in the last statement, which is that the platform basically is able to measure the signaling pathway activity inside a cell, in a disease cell, independent of genetics. You could argue that any disease outside oncology, where you typically do not have recurrent, either single gene mutations or other recurrent kind of genetic context, it's even more important to be able to understand what are the disease-driving mechanisms based on the protein signaling activity, and then try to identify responders and patient and sensitive indications based on that. I&I is a particularly attractive area for us, and we are having an I&I, also an AP3-driven explorer we call I&I Target Explorer that the team has used.
We have identified a number of first-in-class and, if you want, fast follower, best-in-class opportunities and, hopefully, we very soon will announce the initiation of several such programs. We've also assessed a couple for in-licensing. These are cost-efficient. We are accelerating the whole process from target to clinical phase II POC, and that we showed and proved with our internally developed dual WEE1/PKMYT1 inhibitor. We went from the first lead to dosing in humans in 15 months. 18 months later, so in less than three years, we had two oral weekly dosing regimens and have now entered the phase II expansion phase with that.
We predicted that lung cancer would be sensitive, which is extremely unique for WEE1 inhibitors, and we've proven that with many patients out over more than 60 months now, 60 months in heavily pretreated lung cancer patients, single agent, systemic chemo and IO, multiple prior lines, and we're excited about that. That's one of the tumor indications we are expanding into there. But I&I is an interesting area where I think we can really leverage biomarkers. Often it's surrogate biomarker readouts in early phase I, II trials that provide confidence. Biopsy taking is unique. The ease possible often in many of these indications, IBD, EoE, others. And we have human PBMCs. We've developed human PBMCs assays for our oncology assays, so we can look at biomarkers there as well. It's particularly well-suited, and you can't really use genetics. For a number of reasons there, we think it's very attractive.
Now, Peter, if I look at your pipeline, the platform, obviously, you have a lead asset that's in pivotal stage, and then you also are moving into R&I. How should we think about capital allocation across the portfolio?
That's a great question. We've just hired a Chief Business Officer, an experienced such, and our model is the innovation model we really want to move towards. Always was kind of the whole purpose of the company is not to be best at phase III trials. I believe firmly that big pharma is really good at big volume trials. We have a CMO who has been a thought leader, and have a number of approvals in OBGYN cancers. He has led a lot of these, but that's for big volume trials. Our sweet spot is innovation and speed and de-risking off to clinical POC.
Our model will be more and more strategic BD, and that's also why we are starting a number of I&I programs now, where we'll take a compound all the way up to early clinical POC phase II, where really it's still the biggest attrition in our industry in bringing a drug to market, as you might know. So where we want to have the fastest value creation possible with these cycles of phase II. We've been perceived, maybe for understandable reasons, as a binary company with one asset for a long time. It was really the history. We took the highest bar we could. We found an asset to in-license, and we had to start from scratch to prove the platform, and all the seven or eight deliverables have now been proven. Now it's time to really leverage it and deploy it as broadly as possible, as rapidly as possible, as cost and capital efficient as possible.
Okay. I want to move on to your lead asset, ACR-368. That's the first use case, if you will,
Yeah
for AP3, and I know the strategy has evolved quite a bit. Maybe you can tell us why you decided to focus on the serous cell type of endometrial cancer right now, and what data drove that decision, and why do you believe your current phase II-B can support a potential accelerated approval?
Okay. There were three questions there.
Yes.
I'll start with the first one. I mentioned already that we proved actually the OncoSignature, the biomarker approach. Remember that Lilly out-licensing this assay, concluding they could not identify responders. You cannot use genetics. We've proven it works. That was the big thing for us, for investors to prove that. Unfortunately, we are very science data driven, and we did all the blinded prospective assigned preclinical studies that led to FDA aligning with the registrational intent trial, which we did on pretreatment tumor biopsy. All the validation was done on actual pretreatment tumor biopsies, not on archival tissue. In our trial, we continued with that and mandated a pretreatment tumor biopsy. That led to a very slow enrollment, and the reason is very simple.
When oncologists have a drug that you don't need a pretreatment tumor biopsy for versus one you do, and even with sometimes, I hate to say, less effective drugs, with more AEs, they choose the latter because it's faster, easier. Patients are not in need of coming in and lining up for a pretreatment tumor biopsy. Unfortunately, that has really hit us hard with the slow enrollment. That's never good. We used the OncoSignature assay to identify a subgroup of endometrial cancer, which was an identified predicted sensitive tumor types, and we've obviously proven that very robustly, to identify a subgroup of endometrial cancer where we potentially wouldn't need the biomarker, and that ended up being serous endometrial cancer. They are obligate P53 mutated. The biological underpinning support that you depend on CHK1/2 for DNA repair.
We saw in our interim readout, just about 50% response rate across two arms we had run, one single agent, one with a sensitizer, very low dose of gemcitabine, 1% of standard dose. Based on that, we wanted now to move on and prospectively enroll patients, and that's the pre-specified interim readout that's coming in mid-November, we predict, but we say second half Q4 it's going to be, obviously. We are enrolling up to 90 patients. The unmet need is huge. We come in after any approved ADC, after they must have had prior immune checkpoint inhibitor and platinum-based chemo, which is actually the standard of care frontline for endometrial cancer that our CMO had approved a couple of years ago with their dostarlimab and ramucirumab with KEYTRUDA. After that, there's nothing to offer.
It's 13% response rate based on a very large analysis of the KEYNOTE-775, 13% response rate, 3.6 months PFS, and 10 months overall survival after that treatment, and that's based on physician's chemo, and these patients didn't even have IO at the time. That was the pembrolizumab and lenvatinib trial. That's the bar right now. They're emerging ADCs, but we come in after ADCs that are approved and plan to do that. I think, we have had a lower bound of just about 30%, 95% confidence interval. I think 30% with a tight confidence is plenty in that setting. What is happening now is very important.
Our KOLs, and it's non-controversial. Other KOLs are all very clearly agreeing that once you have progressed on immune checkpoint inhibitor and Topoisomerase I inhibitor, which is the payload in ADCs for the most part, you cannot come back and treat again with a TOPO I inhibitor or with ICI.
They become resistant. Pharma now is moving up. There's at least three companies, right? We have Genmab with rinatabart sesutecan. We have AstraZeneca with the B7-H4 ADC. We have the most advanced is actually Merck's, what is it called? TroFuse-005, right?
Yeah.
They haven't shared any data. They have the original phase I data, around 30% response rate or so. They are all moving up. Whoever ends in the front line with that combination creates an even bigger vacuum in second line. We don't talk about lines. We say we come after standard of care, but if you have given ADC and ICI, you cannot give that again. You're left with resistant choice chemo. The 13%, I think, is there to stay, especially longer term. The bar is very low. We think 30% is plenty with tight confidence intervals.
For the interim analysis later this year from Arm 3 and Arm 4, I know you can enroll up to 90 patients.
Right.
Just for this data set, how many patients should we expect? Half, roughly?
The interim is typically going to be around half of that.
Okay.
I also want to say that we have a huge safety database. That was part of the licensing criteria. We have treated, I think, 120 patients now at RP2D, and Lilly had 420 patients across their trials. It's transient mechanism-based heme AEs. The patients don't feel it unless they have one episode, the very few percent that have a febrile neutropenia. So it's really a lab abnormality. There are no irreversible AEs, no non-heme AEs, and nothing of that kind. With that database, and if the response rate holds up, we are very optimistic, based on our interactions with the FDA, that we wouldn't need 90.
Of course, that would require justification with the safety data, but it's a very massive 531 at last count, patients treated RP2D. Many of them over a year. So it's a very substantial safety data set. It's really about the response rate based on prospective enrollment. So it could be less, but we enroll up to 90, and we'll have that completed by end of year, early next year.
Okay. It seems like the bar, Peter, you think would be lower bound 30% response rate from either cohort.
I would say even a target response rate of that. If you separate
Yeah
well above the 13%.
Mm-hmm.
Which is the biggest study there is. That is 218 patients combined or something.
Mm-hmm.
Then you should, with durability that, say, five and a half, six months.
Yeah
You would be in very good shape. I do not even think it needs to be the lower bound at all.
Does the FDA share?
But of course, it depends on whether there comes other approved things.
Yeah
in that, but there's nothing in development. As I said, the ADC ICIs are combining to move front line, and whoever comes first will own that space, I think, personally.
Does the FDA share that view that 30% lower bound response rate would be sufficient to support?
They share the view, also based on our interactions, that it's always a matter of subject to review at the time of submission compared to whatever standard of care there is.
Right.
If the standard of care is where it is now.
Yeah.
Then if you beat that, you beat it. It's very straightforward. It's non-controversial with the single agent in Arm 4 to go for registrational intent. We've also publicly declared that if that data is good enough, even if it's inferior to the ULD desensitizer Arm 3, we will most likely choose that just because of the simplicity of the registrational intent pathway.
Mm-hmm. So you would prefer the monotherapy, which is cohort 4.
It's our drug.
Yes.
Gemcitabine is not ours. Yeah, absolutely.
Yeah.
FDA has agreed there's no single-agent activity of the ultra-low dose of gemcitabine, but they will always ask, is it a meaningful contribution
Yeah
or contribution of components?
Yeah.
We'd have to then justify that with the data. If the single agent holds up, which we think is a driver of activity, then we would any day of the week choose that, I think.
What happens after the interim analysis, assuming one arm or both arms meet that 30% response rate threshold? What do you need to do before you maybe submit or filing, or do you need to generate more data? Walk us through the sequence of events.
It depends on the data.
Yeah.
You could go anywhere from is it just barely good enough data? Prospective enrollment is sometimes tricky. Is it just good enough data? Then we will probably just continue the enrollment and share the data and discuss with the FDA how many patients we actually will need to enroll. If it's stellar, outstanding data, kind of meaning where they are now, you can imagine BTD comes on the table right there, and that's where you discuss already kind of whether you can somehow accelerate, get priority reviews and everything, so you accelerate everything. Just continue and complete the trial. That's how we are going about it.
Peter, can you talk a little bit about confirmatory study? Obviously, you're weighing two options. One would be frontline switch maintenance, and second option would be maybe second line, post-ADC, post-checkpoint inhibitor.
Yeah. We have alignment at high level about the non-controversial switch maintenance with frontline ICI, where in that long maintenance phase of immune checkpoint inhibitor, we would do a plus minus ACR-368. However, for the reason I just told you that more recently it has emerged, the whole competitive landscape moving up towards frontline with ADC combined with ICI, that leaves that massive vacuum and hole in second line. A completely straightforward path would be to take a smaller population and go for serous endometrial cancer in a confirmatory trial, kind of continue the same thing as you did with the accelerated.
That is something that's becoming more and more attractive for all these and other reasons. It's also much smaller patient numbers because you don't need a placebo arm and so forth. It's basically against standard of care, right? That one is becoming very attractive. Serous endometrial cancer is about 50% of mortality and the most aggressive subgroup, despite the low incidence.
To that point, can you talk a little bit about the commercial opportunity here in serous? Obviously high-
Yeah. There's about, in Europe and the U.S., where we are going out with a lot of sites right now and enrolling very fast with a lot of enthusiasm. There's maybe 16,000- 18,000 deaths per year. If you just use the mortality rate, and then the facts stated by KOLs, that pretty much all patients, or most of them progress to second line and third line 100%, there's nothing there. You can anticipate that just using the mortality rate and call it seven months' duration of treatment and your $30,000 per month, you get very significant numbers in a vacuum setting for disease.
Okay. Let's switch over to your second program, dual inhibitor ACR-2316.
Yeah.
Maybe tell us a little bit about your design of this molecule. Why do you think dual inhibition is a good idea?
That is our first poster child that really has opened up for this innovation model of rapid cycles to proof of concept. As I mentioned, we developed that from initially not even the optimized lead to dosing in humans in 15 months, and had in less than three years, proof of concept, with RP2D in lung cancer and are now in phase II, just continuing that in other tumor types. We use the platform, we use co-crystallography and all the traditional state-of-the-art methods, but then we take hits and leads into cells, and we can design and choose molecules based on the pathway effects on top of the traditional in vitro assays. You don't guess, you don't infer from in vitro artificial assays. You see what a drug does inside a cell, and we did that for ACR-2316. The most activated kinases are CDK1/2 and PLK1.
You can see all kinases activated and inhibited with our kinase substrate relationship, where we also use AI to expand that. At the first dose level, we supported the phase I trial with our PBMC assay, human PBMC assay, and we could see at the first dose level, those were the three kinases most upregulated inside a cell. It is a very linear translation into the clinic, and this really paves the way for other assets really quickly. The next one is CDK11, moving very, very fast. We have complete regression in the most aggressive AML models. Again, we did our own subgrouping of AML, independent of FAB classification, and found that subgroup 2, which is very, very high mortality based on some survival data and data that were, is where our drug target access for CDK11 is very activated.
With ACR-2316, which is our dual WEE1/PKMYT1 inhibitor, we have paved the way for this whole model where we move in very fast cycles. Hence, we also got an experienced CBO on board now to really help kind of have bandwidth for these incoming requests to discuss opportunities with our platform and assets. ACR-2316 is active in lung cancer. Obviously, it is the only one causing complete regression in all our preclinical studies against all the benchmarks of WEE1 and PKMYT1 inhibitors in the clinic. Very clean, only neutropenia, transient, no other hemas, no GI, no real fatigue. The only dominant AE is a transient neutropenia at RP2D. We obviously are also including other molecular-defined AP3 identified tumor types, including those that are proven sensitive. Ovarian is included here. ACR-368, we went to endometrial and left ovarian behind because it was not as good in ovarian.
We didn't meet the clinical bar, and there is many treatments there. Patients progress, and our CMO, who's an OBGYN expert, has convinced us that for one that really works there, all these patients keep going to the next line. That's an obvious one for us, endometrial cancer as well, and gastroesophageal junctional cancer also has a very strong rationale. Then lung cancer, squamous and small cell, where we have very long-lasting activity. That one is exciting, and that's our indications. We have single-agent opportunities in second, third line there, although we like to say after approved standard of care. Of course, we have shown for both ACR-368 and ACR-2316 massive synergy with both immune checkpoint inhibitors and tubulin inhibitors consistent with the mechanism of action, and that we can leverage to move more frontline with those. I think that's ACR-2316, kind of.
CDK11 is preclinical. We are aiming for an IND early next year. We have, as I said, complete regression in AML, which is quite unique apparently. There we are planning to. It's active across where the Menin inhibitors are approved, NPM1 mutations, KMT2A rearrangement, rearranged tumors. We also have the myelodysplastic syndrome-derived AML, which is with a lot of splice mutations, which is what CDK11 controls. These are the AML. It's a very broadly active agent, we think, in AML, and are planning just to go all out on AML for that one.
Okay. Thank you so much.
Thank you so much.
That's all the time we have.
Thank you so much.