Okay, it's still good morning, everybody. My name is Chris Shibutani, member of the research team, along with a guy who does all the work, Roger Zhu. The Goldman team is very happy to welcome Exscientia to the 44th Annual Goldman Healthcare Conference. We're excited to have a gentleman who wears many hats, including previously a hat at Goldman Sachs, so he's been an alum- Sure. A recovering alum, Ben Taylor, Chief Financial Officer and Chief Business as well, you? Strategy. Strategy, okay. Kind of more important than that. Dave Hallett as well. Your official title is? Chief Scientific Officer. Oh, very plain. There's, like, all sorts of weird, cool titles that you guys have in your C-suite. No? Right. Okay, we'll have some twists here. Yeah. You have me set up for all sorts of expectations. It's a fascinating time to be you guys, right? Two letters have just sort of, like, hung like stars in the sky, AI, artificial intelligence. Anybody who's sort of invoking technology to do things a little bit differently, theoretically and prospectively, smarter, faster, quicker, cheaper, all sorts of things, foundational ideas, right, that have underpinned the convergence of technology with healthcare. Often it's a clash, a meeting of the minds, disagreement, different kinds of species. You guys have embodied over the course of the corporate history, an evolution of your own point of view, and it starts with your own capabilities, et cetera. Where we're at now is fascinating because as a public company, I think many of the companies in your space, broadly speaking, have some elements of that, and we'll go delve into more detail. Very importantly, there's just the foundational premise of what you could do to be faster, cheaper, smarter, et cetera. How you can invoke your processes into everything that has to do with this difficult journey of discovering and developing drugs and bringing them forward, and then making a decision, which Ben, you'll go into in terms of how do we turn this into a business and with visibility investors can care about. We're here at a healthcare conference, and it's about doing your own thing, eating your own cooking, developing that proprietary pipeline. Everybody who has this kind of approach is taking a slightly different tactic in terms of saying, "We have such a sharp set of tools, we're gonna go after really difficult targets," or, "We're going after validated biology." There's different shapes and stuff. I think it's fascinating, and it's great to have you guys here because of the many moving pieces here with your business, which really start with something that has drawn a very wide investor audience. I think Roger and I talk about the folks who's inbound, it's global. It's not the pure play therapeutics folks; it's often technology, it's CROs. It's just there's a broader audience for your story. Really looking forward to the discussion to getting the update here. With that as a lengthy premise, Ben, perhaps you'd like to tell us a snapshot of sort of how you're feeling about where you're at? Yeah ... at this day here in June. No, that was, that was a terrific intro, Chris. Thank you. As always, you summed up a lot of important things. One thing before we even get started, I think it's really important to demystify AI, because it's this big word that means a lot to a lot of different people. Realistically, it's just a tool. It would be like decades ago of saying, "I'm using a computer." Right? Now what we're doing is, what it allows us to do is take a stream of logic and program it so that it can run on its own, that stream of logic. That's all AI is, in essence. It's not about using AI or not using AI. Everyone will continue to use it more because it's a better tool than a lot of what's out there. It's what you do with it and how you validate it. That's really important. That actually leads back to how we think about our business model, which is, how do we practically apply all of the tools that we're doing? Because we are actually, what we design, most of it does use AI, but we don't have to. What we wanna do is reach a target product profile that's gonna have the most clinical and commercial impact. By doing that, what we do is we start with what's that end goal? What's the differentiated biology, what's the differentiated chemistry that we think is gonna get to it, and then how do we develop a technology platform that gets us there? That's really important when you think about how the company is built up, because one of the big questions that we get is we're doing so many different things. In fact, what those are different projects that had different problems that we had to solve. Now all of those different solutions have become a platform together over the last 15 years. That's how we're really designing the company. It's going through and solving a problem by problem, every problem that we run into in the drug discovery process. What that has allowed us to do now is move five different AI design candidates into clinical trials. We've developed a personalized medicine platform that has actually prospectively been tested in its own clinical trial, and led to better outcomes than the physician's choice. It's how do we get to those end results, those things that could potentially bring benefit to a patient, through the technology that's important, not the technology itself. Right. No, I think that's fascinating. One of the reasons I think the ecosystem supports many different companies is because you are ready for prime time without necessarily having every font for every letter of the alphabet and the full lexicon with which to begin composing poetry, prose, and limericks, right? You can have sort of a positioning and a strength, and then build upon that. Like in the original days of the IPO, it was just like I say, Exscientia, you say small molecule oncology, and I say company A, you know, biologics. You know, from that standpoint, and you have been building your capabilities based upon this tactical approach for, like, incorporating the learning of the problems that you are addressing and hopefully solving, you know. I think that's been very interesting to see. Where are you in that progression where you feel like you can actually stand out there and puff your chest and sort of say, "Yeah, we have actually some, you know, pretty muscular, critical mass here," beyond just the original native, we're really good at this, we're really cool at this, but to stretch further? Yeah. I think there's three things that I look at internally and say, "Are our systems working or not? One is, are we actually putting drugs into the clinic and moving them forward? Correct. Because all of the technology in the world is great, if you're not actually putting a compound forward that could become a drug, you're just theorizing. I think that's a really important part. The second part is we've got a lot of external validation. We now have entered into, and then entered into a second partnership with both BMS and Sanofi h ad drugs in-licensed from them, as well as milestones hit and different things that moved forward with that. That's really important because that says it's not just us defining what validation is. This is actually an external partner that's really doing it. The last part is that we are solving problems that other people haven't been able to solve. We'll probably talk about some of the different drugs, but PKC- theta is actually one of my favorite examples because that was... There are on record 15 different companies that have started a program for a selective PKC- theta inhibitor, and all of them have failed because it's a really tough target. It's a really exciting biological area, but a really tough target. We were able to come up with a solution within about 18 months that was highly potent, highly selective, and had a really nice, clean drug profile that BMS then went and in-licensed. That was really a sign to us that, you know, the application of our technologies is having a really strong effect. Right. Yeah, no, I think you really identified two things. Most investors are very familiar with the standard playbook, the validation. You know, on PowerPoint slide 18 of 24, you see that Bristol, you see that Sanofi logo, right? That's engagement, that's commitment, that's, you know, putting an iron in the fire and representing the fact that they believe in the promise and the potential of really putting minds together and doing some development. Then also, there's the key crucible, which you and a couple other players in this space have matured into in terms of like, we have molecules in the clinic. I think this is actually a really attractive and interesting time as we were to think about the 12 to 24 months ahead, as we begin to turn over some of those clinical cards, right? Ultimately, that's when the healthcare specialists will start to be really attentive and sort of say, it's like: Yeah, I get the premise of how you got here, but let me, you know, test drive you, consumer reports compare you based upon the clinical data. A couple of moments of truth ahead, which is a great segue. The truth seeker, Roger Zhu will now go into the pipeline and interrogate about where we are with all sorts of cool stuff. Go for it, Roger. Great. Thank you, Chris. actually, I want to jump back first for a little bit. you were talking a little bit about how you're incorporating, you know, drug design and patient selection to achieve better outcomes. naturally, I think, you know, this segues into, Can you talk a little bit about your EXALT-1 study? I think it's a little bit underappreciated, and it really sets the foundation for your business. Oh, thanks. Yeah, we love that study, and we've actually started what I'd call a follow-on to that too, which should be really exciting. There's a lot of discussion with clinicians on how we could move this forward more holistically. EXALT-1, for those of you who aren't familiar, was actually a prospective clinical trial where we used our personalized medicine platform to predict patient response in end-stage hematology patients. What we did is we actually took live tumor samples from those patients, put them into a laboratory environment that we've created, tested them against a large panel of drugs. This is really it's not the idea of a cell line or some sort of external model. It's actually using that tumor microenvironment. We've got the immune compartment next to the cancer cells and stroma if there was that in the sample. It really mimics the biology. We can look on an individual cellular level, phenotypically, genotypically, and transcriptomically, how it responds to a drug. From that, we can say, "Okay, in total, this is how this patient's biology is reacting to that sample. The cancer cells are dying or not. Are we killing the immune cells? Are the immune cells transforming, changing, producing something?" From that, we can say, the biology of this patient's tumor is responding, then we can dig into why. What the EXALT-1 study let us do is look at that and say, "I'm going to, out of these," it was about 140 different drugs, "prescribe this drug to this patient based on their individual biology, not because they've got a certain disease, but because I'm seeing that reaction." What we saw was a highly statistically significant improvement in progression-free survival and overall survival with multiple year responses. Really, really exciting both for those patients, but also because it really showed what we're seeing in the lab translated to human biology. Right. You touched upon a little bit of this precision experiment feature of the EXALT-1 study. Could you maybe talk a little bit about how you're applying that to some of your candidates in the pipeline? You're of course, doing it with the A2AR antagonist. Yeah. How about for the 617 program, the CDK7? Yeah. Dave, do you want to answer that? Sure. Just coming back to the A2A project, I think a critical feature of that project was not just the kind of high quality design of the small molecule, but the critical need to identify the patients that would respond. The work that we're doing there at the moment is to how do you identify those patients that kind of meet two criteria? One, their primary tumor has high adenosine levels, or as we call it, the adenosine burden score. Secondly, has an active immune system, because you can't recapitulate that preclinically. You need both of those, because the working hypothesis is that high levels of adenosine in the tumor microenvironment suppress the immune system, and that's kind of why one of the reasons why checkpoint inhibition fails. We've been able to, in a laboratory environment, ex vivo, uncover what we call an adenosine burden score, which kind of correlates with the presence of high adenosine in the tumor microenvironment. That, that gene signature involves is a B cell kind of lineage signal. Pleasingly, we were able to do that ex vivo, is that there's some recent data from other players in the field that have also shown that that probably is a valid signature. They did so after running a clinical trial and after spending kind of tens of millions of dollars. I think we've got a handle on a patient selection strategy for A2A. For CDK7, it's a good example of sort of breadth of the platform we're using there. The idea here is really about, okay, how do you identify patients that are most likely to respond? CDK7 is interesting scientifically in that actually a lot of tumor types respond to CDK7 inhibition. It's an important mechanism. We've actually profiled probably 40 or 50 different tissue samples now, and seen probably 3/4 of those samples respond in some way. The depth of response does change by profile. We published some data in ovarian cancer, where we were able to separate out high grade from low grade serous ovarian cancer. What we see with our CDK7 inhibitor, again, this is using primary human tissue, that high grade serous ovarian cancer responds very well and at relatively low concentrations to our drug, where the kind of the depth of response in low grade, is much less. We can use that to select those patients in the clinic. It's not just about patient selection, it's also about pharmacodynamic biomarkers. We can do that as well. We're in the middle of the moment, we've identified some novel and stable PD biomarkers for CDK7, which we'll talk about in the future, which again, we can use and deploy in the clinic. These are non-invasive, so they will tell us, how deeply we're actually inhibiting the system. Yeah. I think Dave brought up a really important point there. You actually don't want to do this post-hoc in a clinical trial. You actually want to do it during discovery, right? Like, while you're deciding which compound you're going to take forward and how you're going to think about that clinical plan. That's what is really exciting to us, is we're getting to at least the same, if not better, results in a very short period, at a very low cost in a laboratory, rather than several years later after running a, you know, phase II clinical trial. Yeah, that makes a lot of sense. Actually, maybe could you talk a little bit about this simulation-guided trial design ABS scores? Because I think that's, you know, that's ultimately what you're proving out with what you're saying right now. We're really excited about where we're going with development because we wanna have that same sort of data-driven philosophy for our clinical plans that we do on the discovery side. What we do for all of our clinical trials, we start with model-informed clinical trial design, which means you literally simulate, you know, hundreds of thousands or millions of clinical trials with different variables to see what is going to have the most influence on the success or failure of that clinical trial. You can hone in on those areas that you need to get right, that you need to change for inclusion, exclusion, or have the best data for before you ever start the trial. That's before we even lay out the blueprint. What we do along with the clinical trials is basically all of our trials are continuous adaptive trials. This is funny to me because if you look back, and Janet Woodcock published a paper on this, I want to say it was 2017. The FDA and the clinicians are actually wanting better trial design out there, but industry's been relatively slow to adopt. We are fully embracing the adaptive continuous trial design, which allows you to actually get to the answer faster and more fluidly over that process. In correlation with that, adding in the personalized medicine, what we're able to do is actually run that dose escalation portion, for example, while we're also testing a biomarker signature prospectively. They could have different statistical turning points, but you get to a point in the trial where you've determined your dose, but you've also determined, is this biomarker correct or not? Then you could potentially flip that over and make that into an inclusion, exclusion criteria. Let's maybe pivot now into the 546 program, the A2AR antagonist. You're currently running your phase Ib IGNITE study. Maybe you could briefly introduce that to us and talk about the potential indications you could see 546 addressing. Yeah. Dave, do you mind? Sure. Again, one of the critical components of that study is use of this adenosine burden score also needed to help us select which patients are likely to respond. This is not an all-comer study. Through both data we've generated and bioinformatic searches on publicly available data, we identified initial subset. This is renal cell carcinoma, and small cell lung cancer. The reason we've picked those two is that group in particular, has a higher preponderance of adenosine high tumors, and concomitantly, a low Tumour Inflammation Score. I'm sure the audience is familiar with this, is that the work in the PD-1 field, was greatly helped by, there is a gene signature that basically, that indicates basically how hot or cold a tumor is and how likely it's to respond to checkpoint therapy. What we found is that a high adenosine burden score inversely correlates with the Tumour Inflammation Score. Which makes sense because the idea, as I said before, is that adenosine is actively suppressing the immune system. We've coupled with that, this gene signature score, we will use that alongside, and particularly in the dose escalation phase in these two patient subtypes, and look to validate that signature in terms of like patient selection and patient response. Move into a, we'll evaluate the data continuously because it's an adaptive continuous trial design. If we see, performance of that signature that we're actually seeing kind of a signal or a patient response based on that signature, we'll move into a dose expansion phase. How is the patient enrollment going with that study? I think you dosed your first patient just in May. Yep. Yeah, do you foresee any challenges with recruiting or is everything progressing as expected? We're not going to give specific guidance on enrollment, but I'd say we haven't had any problems with it either. Okay, great. It's sort of on schedule. That's good to hear, of course. I guess I just want to dig a little bit into this mechanism of action. You know, you're running this study in combination with the PD-1. What kind of drives that confidence in giving this ability that 546 plus, you know, this PD-1 will kind of reverse immune suppression, if you will? What makes you confident in that concept? I'll come to that in a second. I think one important thing to bear in mind as well is that, so with the first clinical study we did with EXS21546, it was in healthy volunteers. A few people asked us the questions around, well, why are you doing a healthy volunteer study in an oncology indication? That's a little bit unusual. It was very deliberately done, is that the burden on toxicology and data for a healthy volunteer study is higher, so you actually get additional safety data. More importantly, what it's allowed us to do is actually more rapidly get into the combination study, because that was always the plan. I think the body of evidence suggests that A2A as monotherapy is probably limited in use, is that your real target audience is to, again, essentially reveal the activity of the host immune system and allow the checkpoint inhibitors to work. I think the confidence that we've got around why do we think we've got a good chance of this study working? One is the quality of the drug itself. We spent a lot of time and effort understanding the pros and cons of A2A selectivity versus... Because there are a number of kind of sort of dual antagonists out there. Remind the audience that in our hands, the non-selective molecules we found had unusual and kind of sort of negative pharmacology, and that they had bell-shaped dose response curves right in the middle of the clinical concentration range. We think we've designed a molecule that is selective, is highly selective, and will actually address the target. Combined with that is this combination of, that we have this gene signature that we will monitor in patient, and with in a patient cohorts, where we believe that 30%= of those tumors are likely to be adenosine high. We'll identify those. We'll be able to also show at the same time, were those tumors, did they have a low Tumor Inflammation Score? That's the two things we're looking for. These are patients that are relapsed or refractory to checkpoint inhibitors, and did we, did those patients actually recover function and do the checkpoint inhibitors start to work again? You mentioned earlier there's some competitors out there with a dual A2AR, A2BR antagonist. You know, the field recently, there was a study, you know, recently published at ASCO, you know, showing the effects of combined PD-1 TIGIT and this dual A2AR, A2BR antagonist that I was just talking about. This was done in non-small cell lung cancer. Just curious, do you think there's any read-through from that study? I know that the focus of that study was clearly on, you know, what does TIGIT get you with PD-1? You know, there was also this dual A2, you know, A2AR, A2BR. Do you think that carries any read-through to your program? Why or maybe why not? I would say not in that case, because the agent itself is a dual inhibitor. And in our hands, and we've repeated several times, is that, we see at a, at a functional level, this bell-shaped dose response curve. We would question kind of what level of inhibition, functional inhibition is taking place in the clinic. I think the more, the more useful data, is probably from the people like Arcus and iTeos. They themselves have selective A2A antagonists. They're publishing data. I think iTeos are looking in the right place as well, in terms of, I think they've come out and talked about this kind of B cell signature, that they, that they identified kind of post-hoc after a clinical study. I think that the read-through from those two is probably the more useful indicator of A2A as a selective mechanism. Yeah, it's a good example of how you really need to think about a drug in its totality. This is a target where you're gonna need IC80, IC90 times that are very high to be able to get the inhibition level. If your pharmacology is off, you're not even gonna test the thesis. It doesn't matter how potent or selective it is, if you're not there, you're not having the effect that you want. This is actually going back to the fundamentals of why the company was formed, right? We look at a lot of these drugs and say, "Okay, as an industry, we're really good at making potent compounds. We're okay at making selective compounds, but we're really terrible at making clean compounds that don't have a lot of those downstream issues." We see a lot of ADME issues, we see a lot of PK issues. We see, you know, things that are going to cause the drugs not to do what they want for design reasons outside of biology. I was also going to note, you're of course, enriching for your adenosine burden score as well in your studies, which is clearly different as well. Right. That will have other implications. Well, that's an important note. With the with the P1 failures, you typically are going to see an enrichment as well for those higher scores, as the tumor microenvironment changes. Right. J ust maybe one last question on 546. You know, beyond, non-small cell lung cancer and maybe renal cell carcinoma, what other indications do you maybe think about 546 being able to address? I think that will come out of the analysis from this study. I think the key point here is to validate the signage we've seen. Using ex vivo samples is that we have seen adenosine kind of high scores in things like, even like breast cancer. It's a smaller population in terms of, it's like sort of 10, 15% of those samples. What's critical here is like, let's demonstrate and prove that the signature that we've identified works in this particular IGNITE-AI study. If we will then be armed with a signature that we can then use to select patients in broader indications. It's important to note, it's a predisposition. It's not a mutation of the cancer. Yes. What we're actually seeing is on the immune cells, this genetic signature, that when it's in a certain tumor microenvironment, results in that high adenosine level expression. It's a different way to think about. You know, we're always thinking about a genetic variation on a tumor, but actually it's the patient's interaction, that immune system interaction that's important, too. Great. Yeah, maybe let's, you know, believe it or not, there's other, you know, compounds in the clinic, or soon progressing towards the clinic. Maybe let's just talk a little bit about those, given the time. On the, on the 617 program, we talked a little bit about that earlier, but, you know, for those who are not familiar, could you maybe talk a little bit about the program, and then where you've gotten this target product profile today? Potentially, you know, what are the advantages over what else is in the, in the landscape? Yeah. I'll give the quick overview, and Dave can dive into some of the science. CDK7, really exciting target because we think we've seen a lot of the biological validation come out of CDK4/6 already. However, transcription is a very common escape mechanism for CDK4/6 tumors. CDK7 actually has a dual inhibition of cell cycle and transcription. Really exciting biological target. There's been a couple of issues in how you design for it. A lot of the compounds that have been brought up are irreversible, which is gonna cause a lot of toxicity issues. The ones that are reversible, actually have a key issue with that they're substrate, they're transporter substrate. That's going to, and Dave can talk about this in more detail, but lead you to a lot of GI toxicity dosing issues and other things like that. We're moving that forward in the clinic right now. There, it's, we've, we have given guidance that we will announce the first patient moving forward this half of the year. I know that's obviously a pretty short time frame. We're moving that through into the clinic. It will be an adaptive continuous trial. We will talk a little bit more about the indication soon, but we are actually seeing a pretty solid response across a number of high-grade tumors. Because the more pressure that tumor microenvironment is, it seems to respond with transcription more often and be very open to the CDK7. Dave, I don't know. Yeah, I think all I'll add to that is that I think there's an interesting theme with 617, the CDK7 inhibitor, as well as our MALT1 compounds and LSD1 that are in IND-enabling studies, in that the thematic is really around careful thought and then delivery against a target product profile that took into account patient journey and therapeutic index. CDK7 is a really powerful sort of a mechanism. You have to allow the cells time to recover, because if you drive it constantly, it's quite a toxic mechanism. We designed with 617, deliberately designed the molecule that with A, was reversible, so that we could control time on target. More importantly, also thought about human half-life. We didn't want a human half-life that was too long, 'cause we wanted to ideally be probably maybe twice a day dosing would be fine. We uncovered that time on target. If you cover IC80 for about six or seven hours, that's enough to drive efficacy. I think we've come up with a molecule and we've modeled this kind of extensively that we feel maximizes the therapeutic index as we go into the clinic. As Ben points out, one of the competitor compounds, while being very potent and very selective, also has significant transport issues, which we believe will interfere with human pharmacokinetics and also probably drive some of the GI toxicity that they're seeing. With MALT1-LSD1, again, really around MALT1 was the a flaw that we've seen across the entire kind of family of compounds that are in the clinic at the moment. Inhibition of UGT1A1 and the hyperbilirubinemia that goes with that. With LSD1, again, the idea here is being addressing the patient need. Kind of, we've designed a compound that is reversible, is brain penetrant, as an appropriate kind of human half-life. Again, acknowledging the fact that LSD1 has important functions in healthy subjects as well. It goes back to the comment that I think Ben made earlier on about careful thought about the target product profile and who, and the fact you're going to probably be dosing patients both as monotherapy and combinations. I think we're very proud of those assets in the clinic. Yeah. It's, it's funny because, we often say, you know, we try and start with the target product profile. People say, "Well, doesn't everyone start with the target product profile?" The, the answer from a conceptual viewpoint is yes. When you actually get into how it's designed, it's not looking at the entire target product profile. It's sequentially solving product by product in a, or problem by problem in a traditional way. You do potency first, and then from those potent hits, you figure out, can I get some selectivity out of this? If I've got a lead or two that is potent and has some selectivity, then I worry about everything else. Going back to your original statement, Chris, what we're able to do with a better tool and better systems and better process, is actually put those different metrics on par with each other. We can think about clearance time at the same time that we think about potency, at the same time we think about selectivity, at the same time we think about, "I don't want this to be a transporter substrate." That's really important because then you come up with a balanced molecule that is much more likely to be able to test the actual biology than just be something that's potently gonna hit on a target. Great. Thank you. Excuse me. Maybe shifting a little bit gears now. Excuse me. Could you talk a little bit about your PKC theta program and maybe the collaboration and partnership that you did with BMS? You know, how did that collaboration come about? Were these, were the metrics for your target product profile, was that, you know, kind of done in advance, or was that more of a collaborative process? PKC- theta, this is, this was a target that remains of high interest in the immunology and inflammation space. It is a target that I think at least 15 companies tried to get a compound across the line. I think the most advanced was as far as phase II. No company ever managed to kind of address the significant issues with selectivity against other isoforms. We were able to design a molecule that met their desired target profile as very high activity against PKC theta, very broad selectivity, particularly against kind of related isoforms. We were delighted when BMS in-licensed that asset and took it on into healthy human volunteer studies in February of this year. I think that was part, the success there and also the work on LSD1 and MALT1, I think underpinned the decision by BMS. I remember a very good management team to kind of expand into the second collaboration. It's a really hard target, and I know we're almost out of time, because actually the pocket is very similar between the different related receptors. Actually designing something that would be highly specific is incredibly important for this target, because you're talking about a modulator in the immune system, so you don't want to start setting off all of the different PKCs and setting them off in different directions. You're gonna have a lot of downstream impact from that. I think having very challenging objectives and ambitions, but also having a very precise and powerful tool is, I think, a characterization of you amongst the leaders in this space. I think just to put a bow on everything here, you also do have a solid business model and a balance sheet. A balance sheet. Ben, talk to us a little bit about just the financial situation, so that we have confidence that, you know, the lights are staying on? Sure. For all of these projects as a close for us. I'd love to close on that. We've got at the last quarter over $550 million in the bank. Our burn, we expect that to provide us with several years of cash runway without any additional financings or partnerships. If you look at our numbers from the last quarter, there are a number of one-time capital and change in working capital charges in there. If you back out that, you'd be a little under $50 million-ish for a cash burn, and that's probably. We don't need a lot of scaling to continue executing on our plan t he clinical costs won't ramp dramatically for really several years if we decide to go into late-stage clinical development. We've got the organization we need. We have the scale we need. That gives us a lot of flexibility and runway in how we execute on the model. We still continue to bring in a lot of cash from our partnerships as well. Right. We've given guidance. We expect, we could bring in several hundred million in cash inflows from our existing partnerships over the next few years. As well as, that we expect to enter into two new partnerships this year. Okay, excellent. Ben Taylor, Dave Hallett, thank you very much. Great. Watch this space. Thank you. Appreciate the thoughtful comments. Thank you. Thank you very much. Appreciate it.