Far side chat with Evaxion. We have, I'm Suomit Roy, Head of Healthcare Research at Jones. And with us today, we have CEO of Evaxion, Christian Kanstrup. This is a very appropriate venue for AI-driven, and we have a lot of talk of AI today. Christian, Evaxion is developing vaccine using the AI platform. As we all are learning about the incorporation of AI for drug discovery, could you just briefly tell us about Evaxion and what is your focus?
No, I'll be happy to. First of all, great to be here. Thanks a lot for the invitation. No, in essence, what Evaxion does is we are an AI-driven company focusing on developing precision medicines. We have an AI platform, which we call AI Immunology, which we use for discovering and developing novel candidates within cancer and infectious diseases. I think an interesting thing about Evaxion is we were actually founded back in 2008 as an AI-first company. The two co-founders, they wanted to use machine learning to decode the human immune system to develop novel medicines. Today we have evolved into a clinical stage tech bio company with a broad pipeline. We heard about the tech bio concept from the keynote speaker this morning. That is actually what we have been positioning ourselves as for a number of years.
I think the fact that we were founded in 2008 also means that we have had a significant head start compared to many of the newer AI companies out there today, allowing us both to develop our platform, but also, importantly, validate it.
One of the questions which always comes up, and essentially AI is a black box to most people. We had 20, 30 plus years of high-throughput way of looking for either designing peptide or looking for structures in small molecules and trying to identify 10,000, bring it down to like 20, then bring it down to like 10 in the animal and one to the clinic. How does the AI aspect of it accelerate the process? Where is the power?
No, first of all, this is about the black box. I would say that we also hear when we are discussing with potential partners, I mean, it is a black box. How do we know it works? I would say very clearly that it is clear that AI has transformed drug discovery, both in terms of drastically speeding up the process, also improving the precision of how we discover and develop new candidates. Instead of relying on, you can say, time-consuming lab work, we now have a significantly different speed compared to what we had previously. I would say the difference compared to the high throughput is a much higher sensitivity we see with, and specificity we see with the AI approach to discovering novel drugs.
If you have to quantitate like the time saved versus prior approaches, new antigen identification, it was like, I think if I have to guess, like two and a half years, something like that. How much has it reduced down to? Is it just the pace of it, or do you see the quality of the antigens it's predicting are more immunogenic? Where do you see it?
It's actually both. I would say now, if we look at the new antigens, I think that's probably around two and a half years you're saving. Also, if we look at infectious diseases where you have a proven approach like reverse vaccinology, which typically takes three years, we can deploy our AI platform, and just in 24 hours, we can get a ranked list of novel antigens.
Within 24 hours, you'll get it. Wow.
Of course, that's a significant time saving. On top of that, we also get this ranked list of potential new targets, i.e., we can just select the top ranked ones. Hence, in terms of testing, we can go by testing significantly fewer targets than what you would have to do if you were doing it in a traditional way.
Do you think there is a marked difference? Because new antigens are hard to predict.
Yeah.
Versus bacterial, viral, they're almost epitopes unknown for a long time. These are the highest immunogenic epitopes because from the sequencing and all. Do you see this is more effective in predicting the new antigen side versus bacterial epitopes? Only a few things you can change there.
I would say, of course, given that if you look at the tumor mutation, the neoantigens are patient-specific, right? That also, of course, means that you cannot have a large database where you can use that to predict the right neoantigen. You need to do that on a patient-by-patient basis, which of course requires a swift and rapid approach to do that in an effective way. That is where the AI approach really makes a difference. I would say also for the infectious diseases, it is clear that not only does it increase the speed, but it also allows us to discover novel targets that you have not been capable of discovering previously. Take, for instance, our EVX-B2, which is a gonorrhea candidate.
It is still in preclinical development, though, but now the field has for the past 50 years been trying to come up with an effective vaccine against gonorrhea. Here we have in preclinical models shown that we can, in 50 different gonorrhea strains, completely eradicate the bacteria.
It picks up those glycolipids of the gram-positive.
Yeah.
I didn't know that. That's really interesting. Talk to us about a little bit on the training the AI software itself, like the database when you train, have you proprietarily developed it or you took it from other institutions, cancer research?
It's a mix of different things. We have, I would say, three sets of data. We have what I would say proprietary data, which are data we have been generating ourselves, both preclinical data, but also we have been conducting three different clinical trials with our personalized cancer vaccines, generating a lot of data, which we have been feeding into the model. So that's one set of data. Then we have semi-proprietary data where we are collaborating with hospitals, universities in order to get access to data, which are not in the public domain. Finally, of course, we are using publicly available data. I would say one thing that's important also is it's not just about the quantity of data, but it's very much so about the quality of data as well.
That is where having this mix of different data sets, which you can use to further refine and enhance your platform, is very important.
This might be a little unfair question to you, but where is there room for improvement on your database itself? If we have to compare yours versus, let's say, Moderna's and BioNTech's, is it just that they have access to a larger patient sample, they have a better trained model, or is it like that is not where the delta lies? That delta lies in something else?
I would actually claim if you look at data that's available, that we have a much higher predictive ability of our model than both Moderna and BioNTech. In our Evaxion phase II trial, we published the one-year data in the fall of last year. There we saw that 79% of the new antigens were eliciting an immune response. If you look at Moderna, they have not yet published data from their completed phase II trial with their personalized cancer vaccine, but they have published, I think it was Nature, data on their phase I trial, which is a basket trial where they showed that only 25% of the new antigens elicited an immune response. That's where we have published last year that 79% of the identified new antigens actually elicited an immune response.
BioNTech, they showed in their phase I- B trial that eight out of 16 patients saw an immune response. Hence, it points at, of course, it's always difficult making these cross-trial comparisons, but it does point at the predictive capabilities of our AI Immunology platform is very high. That also comes from this continuous refinement. In the phase I trial with our EVX-01 , we saw that 58% of the neoantigens elicited an immune response. By refining and further training the model, we got that to 79% in the data we released last year.
Probably this speaks to the fact that you guys just did a big deal with Merck and they took like 20% position in the company. Give us a little background on how that conversation went and what was their main focus as they're because they're one of the biggest pharma in the world.
No, I mean, we have multiple collaborations with Merck, and it's true they are our biggest shareholder. They have a 20% ownership of the company. We also, in September last year, entered into a large optional licensing agreement covering two of our infectious disease candidates. That actually started out in, that was in 2023, where we were approached by Merck saying, "Hey guys, we would really like to utilize your AI Immunology platform to develop a novel bacterial vaccine candidate within an area where there are no vaccines available today, but a high unmet need." It has not been disclosed what the pathogen is, but it's one where there are no vaccines available today. That collaboration we started in September 2023. They invested into us first time in December 2023, and they have actually participated in our three last financing rounds.
In September last year, we entered into an optional license agreement, both around the original asset, but also they actually included our EVX-B2 , our gonorrhea candidate in that collaboration. Very pleased around having a world leader in vaccine development and commercialization, not only partnering around one asset, but actually two assets. On top of that, taking a 20% ownership stake in the company. Very pleased with that collaboration. I think their interest was driven by exactly the fact that we can discover novel antigens, which you can't discover with traditional methods. We can also do it in a fast way. It was the predictive capabilities of the platform that was attracting the interest. We actually also have a supply agreement with Merck. They are supplying Keytruda for our EVX-01 personalized cancer vaccine trial, but that's just a supply agreement.
It's much broader, the collaboration around the infectious disease side of things.
Merck has an option trigger second half of this year on the gonorrhea.
Yeah, yeah. The agreement we entered into last year, we got $3.2 million upfront. If they exercise the option to both candidates in the second half of this year, we are set to get an additional $10 million plus future milestones and royalties, of course.
This is what I'm trying to understand. Like why isn't Merck developing their own internal AI just like every large pharma is doing? Second is, how deep is your relationship? Can you, because they probably have one of the largest cancer tissue repository and data bank, can you train your model on their dataset and do you have that kind of agreement?
That's not part of the agreement. The agreement now is focusing on the infectious disease candidates that we have in scope. You can say, why aren't they just doing it themselves? We need to remember we have actually been spending 15 years in terms of developing our platform and also you can say maturing it and constantly refining it to increase the predictive capabilities. It is something that takes a lot of time. I think now I've been working with big pharma myself for many years. I think focus from an AI point of view in pharma is probably more into the whole clinical trial area. How do you utilize AI to optimize those processes rather than it's the early drug discovery?
Right. I understand. Your lead asset EVX-01 in phase II trial in advanced melanoma in combo with Keytruda. Do you get the question where you have to really show how it is differentiated versus any other personal vaccine? Is it just the response rate or is it something else you see that looks better versus other?
There is a lot of focus on, of course, the response rate. That's one thing. I am pleased also to say that in the phase II one-year data we presented last year, we actually showed a 69% overall response rate. If you compare that to the KEYNOTE-006, the Keytruda registration trial, there they showed a 33% overall response rate. There is also a lot of focus on how many of the neoantigens are actually eliciting an immune response, i.e., the 79%. There is focus on, you can say, the durability of the response.
That is also why we actually just early on this year announced that we are extending the trial with another year, not because we have not seen good data, on the contrary, but because we have seen so exciting data and are curious to see how long is the durability of the immune response actually.
The other hot button issue is selecting the indication itself.
Yeah.
Melanoma for a while has turned into a proof of concept indication versus you have to go show it in lung, in CRC if you can, or some other indication that makes the model work. How are you thinking?
I fully agree with you that melanoma was chosen due to a number of different reasons. First of all, it was actually developed as part of a consortium. The original thought was around a consortium with a Danish research institute. That is where you can say focus was naturally on melanoma, also because you have a very high tumor mutational burden here, hence a lot of new antigens to potentially new antigens to select from. It is also clear that the whole concept has broad applicability in other solid tumors. I think right now our focus is on establishing the clear proof of concept in melanoma. Our focus is also not on the later stage, large scale phase III trials. We will do that together with a partner.
Hence, it will be also up for discussion with a potential partner, which is the next indication you want to bring it into. It has a broad potential in look across solid tumors where you have high tumor mutational burdens.
What comes to mind first?
I think we have been looking at a lot of different things, but we have not really made a decision yet because again, we can have our own picks. I mean, I think we have also seen some of the indications are more challenging than others. It needs to.
CRC, you have quite a pleiotropic situation, but CRC is really hard to crack.
Yeah, yeah.
On the other hand, lung, you have a lot of mutations, but you have targeted medicine for each of those.
No, we have actually been discussing lung quite a lot. Even though you have a lot of mutations, it also comes with other challenges. I think for now, focus is on continued generation of great data within the current phase II trial. In parallel with that, be preparing for taking that discussion with a potential partner. We are, of course, also looking at what's the clinical path forward from completion of the phase II trial, how do we as quickly as possible get to a potential registration.
What's your goal for the valuation of the company? Is it prove the technology in one indication, like let's say melanoma works well, is applicable, will get adopted, long duration of response, and then sell the technology platform? Or do you want to use the platform and become a therapeutic company and just keep developing drugs for different indications?
More the latter. I would say we do not have the ambition of becoming a commercial scale company. We have the ambition of doing what we do best, which is deploying the AI Immunology platform, doing the preclinical and early clinical development, and then, of course, bringing novel concepts to the market. We are also within the cancer area. We are working on a precision cancer vaccine concept where we have been discovering a novel source of potential cancer targets, these endogenous retroviral, where we are actually having discovered that these are shared across patients. Hence, we can make precision medicines for a group of patients. The advantage here is that is actually found also in cancer type with a low tumor mutational burden. Hence, you have the opportunity of expanding the use of immunotherapy.
ACR, you are presenting some data?
Yes.
What should we expect? It's a more biomarker or like?
It's more biomarker. We have been doing a lot of immune data analysis. Among others, we have had further data, of course. We will be presented updated data on the hit rate for the new antigens. On top of that, we have been doing a lot of exciting further immune data analysis, which we will be presenting. We are looking very much forward to that.
In a couple of weeks.
Yeah.
You have also guided that the two-year readout for the phase II is going to come out second half of this year. What should we expect as the benchmark that gives you confidence, you know, we should push it towards the pivotal trial?
I would say given the data that we have seen so far, we already have a lot of confidence in EVX-01 . Of course, what's going to be important is to see the durability of the response. And then, of course, if we can continue to see an even further increase in the overall response rate, and also looking at, you can say, the number of converters and comparing that to the checkpoint inhibitor data in order to further solidify that the strong data we are seeing are generated by EVX-01 on top of the checkpoint inhibitor. That's, of course, always a challenge when you have a single arm study, what's driving the good clinical response. That is what we are looking forward to seeing when we get to see the phase II, two-year phase II data, which is coming out in the second half of this year.
For context and response rate-wise, Keytruda alone gives you around 30%-33% response rate, and you are seeing 79% response rate.
We're 69.
69.
69.
69% response rate. In terms of PFS, is it Keytruda only gives you like five months, five and a half months?
Yeah.
Something like that. We want to see anything above that to be additive. Remind me also in terms of the clinical trial sites, are you primarily in Europe or you are?
It's conducted in Europe and in Australia. We have two sites in Europe and two sites in Australia. Yeah.
What's the plan of expanding it to coming to the U.S., also get the India?
Yeah, we actually have fast track designation in the U.S.. We will be now entertaining discussions with the FDA as well in order to guide the design of how would a pivotal trial look.
It would be so most likely after we get the second half long-term data, then those further conversations will start.
Yeah. We are preparing the conversations now also because, I mean, our intention, as I said, is not to bring it into a pivotal trial ourselves, but to do so with a partner. Of course, coming into partnering discussions with a clear view as to what could be feasible from a clinical trial point of view is a strong advantage.
This is great. If you have any concluding remarks.
No, I would just say that, I mean, it is clear that the AI and use of AI in drug discovery and development truly is transforming the way we develop novel medicines. I think one thing is, of course, speed, but more important so is that it allows us to do things that we couldn't do before. I mean, having a 79% hit rate in terms of selection of new antigens in a personalized cancer vaccine, those are amazing data, which really gives me the hope that we have the opportunity of doing something about the still way too high unmet need within cancer.
Also within infectious diseases, having the ability of being able to bring novel vaccines to the market where the field has tried for many years to discover vaccine candidates and combining that with the ever-growing resistance towards antibiotics, it's clear there is a significant unmet need here and that we have the opportunity of addressing via the use of AI. It is truly transforming the way we develop new medicines. I'm very excited about that.
Thank you so much for coming and really good luck.
Thank you.
Great to be here. Thank you.