Exscientia PLC (EXAI) 2022 Q3 法說會逐字稿

Hello, everyone. My name is Brent, and I will be your conference operator today. At this time, I would like to welcome everyone to Exscientia's Business Update Call for the Third Quarter of 2022. (Operator Instructions). At this time, I'd like to introduce Sara Sherman, Vice President of Investor Relations. Sara, you may begin.

Thank you, operator. A press release and 6-K were issued this morning with our third quarter 2022 financial results and business updates. These documents can be found on our website at www.investors.exscientia.ai, along with the presentation for today's webcast. Before we begin, I'd like to remind you that we may make forward-looking statements on our call. These may include statements about our projected growth, revenue, business models, preclinical and clinical results and business performance. Actual results may differ materially from those indicated by these statements. Unless required by law, Exscientia does not undertake any obligation to update these statements regarding the future or to confirm these statements in relation to actual results.

On today's call, I'm joined by Andrew Hopkins, Chief Executive Officer; and Ben Taylor, CFO and Chief Strategy Officer; Dave Hallett, Chief Operating Officer; Mike Krams, Chief Quantitative Medicine Officer; and Garry Pairaudeau, Chief Technology Officer, will also be available for the Q&A session. And with that, I will now turn the call over to Andrew.

Thank you, Sara. 2022 has been a busy year, and we're excited to update you today on our recent progress. We believe that our Patient-first AI strategy in combination with our world-class team and innovative platforms, places us in a position of strength as we create value across our internal and partnered portfolio. We are well capitalized with $625 million in cash at the end of the quarter. This provides us with several years' runway to advance our near-term programs, whilst we also invest in tandem for long-term growth.

Within our platform and pipeline developments, we have made meaningful progress throughout the year, which has continued to validate the technology that Exscientia stands on. Today, we provide more insight as we get it closer to treating the right patients with the right drug in a clinical setting. We are also excited to highlight data demonstrating how we plan to innovate and increase the probability to access to the clinic and similar to what we've already achieved in discovery.

Following our topline data early this year for EXS-21546 or '546 our A2A receptor antagonist, we remain on track to initiate a Phase 1b/2 study by the end of the year. Importantly, our A2A program not only learns about compound, but also allows us to continue to build our confidence in our translational signature on how best to enrich for patients who are more likely to benefit from this treatment. We have also made exciting progress with GTAEXS-617 or '617 our CDK7 inhibitor, new data we presented at the ENA Annual Symposium in October, demonstrating the potential of our precision oncology platform to assess targeted therapies and defined patient populations that are more likely to respond to our therapy.

This also highlights the potential of '617 to be the less cytotoxic compared to other therapies, such as CDK4/6 inhibitors. We believe from our platform, which combines deep learning, functional models and multi model omics data, from our experimental platform of relevant primary human samples, not only helps us better understand the potency and activity of '617, but will also lead to improved patient selection and therefore better patient outcomes. We are on track to file a clinical trial application or CTA for this program by the end of 2022 and expect to initiate a Phase 1 clinical trial in multiple solid tumor indications in the first half of 2023.

In our partnered programs, we are adding an additional oncology target into our Sanofi collaboration. More generally, we look forward to show more details on our pipeline, including additional drug candidates progressing through IND-enabling studies. We recently shared a couple of exciting announcements, including a new strategic collaboration with the University of Texas, MD Anderson Cancer Center to develop small molecule therapies in oncology and our expansion into biologics design. We will talk more about it very shortly.

As our clinical pipeline grows, I'll now hand over to Ben to walk you through how we plan to be as innovative in development as we have been in discovery.

Thank you, Andrew. Looking at the lifecycle of a drug, all of its future potential is determined during the initial designed phase. However, the validation of that potential only occurs during clinical testing. Our strategic goal is to shift the economic curve of the pharmaceutical industry by improving probability of success, accelerating development cycles and reducing costs. As you will see, nearly all of the tools that we use for drug discovery also apply to more effective clinical design and execution. I'm going to spend a minute covering our clinical strategy before Andrew discusses recent advancements in precision medicine, but it's important to understand that the 2 are closely linked. When we use real patient samples in our precision medicine platform to guide molecular design, we are also defining our future clinical programs the phenotypic, genotypic and transcriptional signatures that we identify with our development candidates are the same profiles that we will target in our clinical programs.

We have seen this to be very effective in single-gene loss of function mutations, and our goal is to apply that same targeting strategy to more complex biology. We have given previous guidance that we expect to initiate the next phase of our A2A molecule '546 by year-end, and our CDK7 compound '617 will follow shortly thereafter. Looking forward, we expect that we will have at least 3 compounds that are in clinical development by 2023 and 4 by 2024. This only includes compounds where we maintain significant economic interest and excludes the existing clinical programs with Dainippon Sumitomo Pharma.

On Slide 7, you will see how the strategy translates into practice. When we launched the trials for '546 and '617, we will provide you details on their individual trial designs. But this slide outlines what you can expect for all of our internal oncology programs. We start with what we call ex vivo clinical trials using our precision medicine platform. We evaluate the drug in a multitude of heterogeneous patient samples with in-depth profiling to understand why some of the samples respond better than others. This provides us with a biomarker that we believe will correspond with the most sensitive patients for a specific drug. We can also perform multiple iterations of these prospective ex vivo trials to create a statistically significant sample set before initiating an expensive and lengthy in vivo clinical trial.

Next, we start with an initial clinical trial in a patient population where we expect a higher incidence of the response signature. While the trial is validating the compound characteristics, including safety and preliminary efficacy, we are also evaluating all of the responses prospectively with our precision medicine platform to validate the biomarker. Once confidence in our biomarker signature has been reached at a predefined level, we can apply that biomarker signature to directly enrich the patient population in confirmatory trials.

All of this is with a specific goal of increasing the probability of success in clinical development. We believe that many clinical trials fail because they are not treating the right patients. Precision medicine needs to be incorporated into clinical progression so that you not only define a drug's properties but also who should receive it. In addition, we want to make the trial design as efficient as possible. In our industry, we often see novel products tested using clinical methods that are decades old even when newer methods have been proven more efficient and accepted by regulators.

Our Chief Quantitative Medicine Officer, Dr. Mike Krams, oversaw clinical innovation for all of J&J's portfolio and now make sure that we are adopting state-of-the-art clinical strategy. For example, we utilize simulation-guided clinical trial design to virtually identify the best statistical models as well as critical variables that may influence the success of a trial. Another way for us to advance our clinical capabilities is to partner with a world leader in clinical scientists. The University of Texas MD Anderson is not only one of the top institutions for research and medicines, they also pioneered many of the adaptive clinical trials that now define intelligent trial design.

With this partnership, we aim to combine our expertise in AI-based discovery and development with their extensive experience in oncology and clinical trial execution. Hopefully, the collaboration will lead to both Discovery novel targets as well as the next generation in innovative clinical design.

I'll now turn the call over to Andrew to talk a bit more about our progress with patient-centered precision medicine.

Thanks, Ben. We'll now take a few minutes to walk through 2 of our ongoing programs and how we are doing a step-by-step perspective ex vivo analysis of patients prior to entering clinical trials, highlighted important work for our '546 our A2A receptor antagonist; and '617, our CDK7 inhibitor.

Now on to Slide 10. For background, we know that other A2A candidates in clinical development have seen some albeit limited clinical success with overall low response rates. We believe our clinical success can be improved for enriching and targeting for the right patients. The key to success in the A2A field, we believe is matching the drug to the right patients.

We now have identified a preliminary gene transcript signature, measuring ex vivo adenosine burden in primary samples which we believe could be predictive for '546 response. Our '546 was form signature or adenosine burden score, which we refer to as ABS. In the heat map here, you can see ABS in 43 sample patients from 6 different tumor types. The signature is made up of 6 genes, shown on the slide and 2 normalized genes, which are not shown. This gene signature was initially identified with differential expression of genes in primary mononuclear immune cells after ex vivo perturbation with stabilized adenosine.

The ABS is currently set relative to biomarker levels in unexposed healthy immune cells. The more we increase adenosine seen in the yellow box here, the more adenosine driven immune suppression (inaudible) the potential for '546 to have a real effect on blocking adenosine detection. We see that a number of samples across tumor types of high scores, and we believe this signature is likely applicable across several cancer types as it is based on the immune system.

The next 2 slides will demonstrate how ABS we modulated ex-vivo and how it relates to a validated tumor inflammation signature. This work also revealed potential PD biomarkers beyond (inaudible). We are working to confirm those. This is really important and differentiated and how to think about designing clinical trials to maximize probability of success for these patients.

Here on Slide 11. On the left, we can see here that our ABS behaves more consistently and as expected, in the presence of stabilized adenosine than the adenosine signature from (inaudible) published in Cancer Discovery in 2020. The score can be seen here from transcriptomic studies conducted on 4 primary mononuclear samples stimulated with stabilized adenosine in the presence of T-cell activation. As we increase ex vivo stabilize adenosine and takeaway '546, there is a progressive increase in our adenosine signature score as shown in the orange box. On the very same data set, our signature outperforms other public signatures in terms of specificity and sensitivity to detect adenosine rich microenvironments.

On the right, we show that signature allows the detection of adenosine rich in microenvironment as presented with primary lung and RCC cancer patient samples exposed to increasing doses of stabilized adenosine. We believe that this ultimately shows that the immune response can be controlled by treating primary patient samples with '546 as confirmed in the T-cell activated blood mononuclear cells. We are continuing to collect adenosine and '546 specific data and look forward to sharing more about how we expect to enrich for patients in the clinic. The adenosine gene signature that we have developed will be further validated alongside our planned Phase 1b/2 study starting soon.

To take us first on Slide 12, as we move towards validation, we've identified a relationship between our adenosine signature and important published score for inflammation and checkpoint immune response. We see that as we increase adenosine, we are also decreasing the TIS or tumor inflammation signature. On the chart, we show how we can increase confidence in the score by mining public data sets such as a TCGA database, the NCI's cancer genomics program. For background, we already know from published research that adenosine is an alternative way by which tumors can escape the immune system even if treated with an anti-PD-1.

And indeed, we find that the most predicted high adenosine cases are amongst patients that are least likely to respond to recurrent immunotherapy according to the TIS. This helps us understand potential responders to treatment with '546 in combination with checkpoint inhibition. Our hypothesis is that disease progression under PD-1 therapy may be associated with local immune suppression mediated by high adenosine levels in the tumor microenvironment for a significant proportion of patients as highlighted by these specific tumor types.

This idea is further sustained by the data presented here on Slide 13 in our lung cancer patient tissue cohort exposed to increasing doses of adenosine. We can observe the trends for TIS reduction on the same samples, notably of highest doses of adenosine, underlying further the potential connection between adenosine-rich environment and checkpoint inhibitor responses. Ultimately, this reveals potential implications from a clinic in terms of study '546, in combination with a checkpoint inhibitor. Bringing it back to the example of '546 we mentioned above, we have seen other signatures are inaccurate which means that it's very challenging to identify which patients may response to therapy, and this has been evident from data observed from current clinical candidates. The data we've just shown highlights a highly stable biomarker behavior and our platform performs with ex vivo. And we've taken a similar approach with '617 or CDK7 inhibitor, as we prepare for the clinic.

On Slide 14. As mentioned earlier, we've recently presented data at the ENA meeting of our functional precision medicine platform, combined in single cell sequencing and transcriptomics of disease-relevant primary patient material to maximize understanding of the potential '617 effect, aiming to select patients, we believe will be most likely to benefit from treatment.

For the first time, we've shown how we apply multimodal machine running to integrate data from primary human tissue samples and multiomic sequencing capabilities to predict tumor efficacy of '617. Here, we highlight some of that important data, including data shown at '617 induces less cell death of immune cells than other investigational CDK7 inhibitors, potentially indicating the differentiated clinical safety profile. Using our deep learn empowered high-content imaging platform, we previously confirmed '617's activity in primary human samples with example data in ovarian cancer samples shown at the AACR earlier this year. These expanded results that you can see here on the right-hand side of the slide in the waterfall plot, led us to generally define 2 groups of patients samples, effectively high and low responder groups in the indications tested.

The waterfall plot highlights the ability of our platform to identify from primary human tissue samples, responders and nonresponders to a CDK7 inhibitor across tumor types, which is really excited. It's important to note that this is not showing a percentage of response rate, but it's helping us to understand who the right patients the study are, and we can see that the type of cancer is not a determinant factor in the patient tissue response to CDK7.

Moving on to Slide 15, then using transcriptomics of the same patient samples, we have identified a gene signature, the first established in ovarian cancer for which can potentially predict an outcome for patients, we are actively working to expand the signature of the tumor types. Here on the left, we're showing the gene expression levels of genes that make up our CDK7 predictive signature in an ovarian cancer patient cohort using 30 samples as a training set for a statistical model. We are currently undertaken single-cell ex vivo functional screening combined with transcriptomics after CDK7 perturbation in disease-relevant primary human cancer samples to refine and improve this gene signature further.

We are aiming to use this baseline transcript data combined with a functional assessment of '617 in primary patient samples to model and then predict patient response to CDK7 inhibition. We have currently added more indications to this model and further confirming the model biologically ahead of the trial for validation alongside the trial. So what we are doing is essentially running a clinical trial ex vivo to understand which patients will respond based on the new biomarker prior to enter in the clinic. Therefore, increasing the chances of treatment response and trial success.

As a reminder, the functional layer is from the same technology that we clinically validated in the separate prospective trial of X Old One in hematological cancers. In that trial, we showed that the AI-driven platform was able to select the right treatment for an individual patient, significantly improving their progression-free survival.

Finally, which is not shown here, but was presented on the EMA from single cell transcriptomics data collected after treatment of a samples of CDK7 inhibitor, we have validated existed and defined in novel CDK7 specific pharmacodynamic biomarkers that may enable us to track '617's activity, potentially non-invasively doing our planned clinical study by combining AI-based patient tissue analysis and transcriptomics data we identified both a PD biomarker and the gene signature specific to CDK7 and the drug candidate '617, which may distinguish patients that would be optimally responsive to the therapy.

This (inaudible) using disease-relevant patient tissue preclinically to predict the response of cancer cells to our CDK7 inhibitor means that we have taken huge strides towards the overall goal of being able to deliver the right drug to provide patients. We plan to validate this data retrospectively alongside our planned future trial. Our goal is to make precision medicine a reality for patients for the application of AI.

Before I turn over to Ben to cover financials, I'd like to highlight another key development, the exciting progress with biologics. We're now expanding into generative AI design of novel antibodies. We've taken the potential of AI combined with automated experimentation to build a process for biologic design that we believe is more efficient and will lead to better results for patients. This expansion effectively doubles the addressable target universe of our precision medicine platform, allowing us to develop the most effective drug for patients regardless of modality.

How do we actually define how biologics is developed, and it is by design, not discovery. On Slide 17 today, the current process is driven by experimentally discovering binders, which have many limitations. By virtually generating and designed in precision-engineered polyhuman biologics, we could overcome these challenges. We believe that our method can explore much broader target universe and create a right antibody for a specific target.

Moving to Slide 18. We've already been able to show that our virtual screening methodology of antibodies is now over 3x more accurate than the published state-of-art. Additionally, we can produce accurate protein modeling for antibodies up to 35x faster than our result. We've also generated data in new ways when analyzed using AI will allow more complex and complete understanding of human antibody biology.

Our new automation laboratory will be up and running next year, and we'll allow rapid assessment of essential qualities like affinity, immunogenicity, aggregation and stability that feed directly back into our AI models. By combining this knowledge of the ability to generate complete novel antibody designs virtually, we aim to create better biotherapeutics faster and more efficiently. And as we mentioned last quarter, we've already used our precision management platform to help select patients of biologic programs in our current collaboration with Sanofi.

I'll now turn it over to Ben to cover our financial highlights.

Thank you, Andrew. I'll now take a minute to close with highlights from our financial results for the first 9 months of 2022. Full results are detailed in our press release and Form 6-K. I'll review the results using U.S. dollar using the September 30 constant currency exchange rate from pounds to dollars of 1.11. If you are converting our financials on periodic U.S. dollar to GBP spot rates over the last year, there may be volatility in U.S. dollar consolidated figures. However, it's important to note that we hold our cash balances in either British pounds or U.S. dollar based on expected expenditures in that specific currency.

Therefore, the actual operating impact of market exchange rate movements are greatly reduced or eliminated. Our cash inflows from collaborations for the 9 months ended September 30, 2022, were $117 million as compared to $67.5 million in the first 9 months of 2021. We continue to expect cash inflows from collaborations to remain lumpy around milestones and business development. For the first 9 months of 2022, net operating cash outflows were $15 million in comparison to net operating cash inflows of $8.3 million in the first 9 months of 2021. This has been an important year of growth with investments in our platform, including precision medicine, automation and biologics. We will continue to pursue a strategy that balances near-term monetization through partnerships with building a truly differentiated technology platform and pipeline.

This is why we can feel confident of growing the business while also protecting our cash runway. This year, we have had a number of onetime capital expenditures associated primarily with our automation laboratories and facilities expansion, combining CapEx with our operating cash flows and a few miscellaneous items our net cash burn for the first 9 months was $37 million. We ended the quarter with approximately $625 million in cash, equivalents and bank deposits. We believe this gives us several years of cash runway. And with that, I will turn the call back to Andrew.

Thank you, Ben. Today, we walked you through a few examples of how we are working to transform the industry not just by bringing our AI-driven drug discovery platform into new modalities such as biologics, but also modernizing the way we select patient populations that may respond to new therapies, ultimately allowing us to design better clinical trials. And with that, we will open up the call for questions and answers.

(Operator Instructions) Your first question is from the line of Michael Ryskin with Bank of America.

Great. First, I want to start on the biologics focus, really interesting technology and -- so it's not -- could you talk a little bit more about how you see that ramping overtime in terms of incremental investment going forward between the small molecule platform versus large molecule? And just in general, how should we think about that part of the pipeline evolving? Is there any particular target for you for internal in terms of small molecule, large molecule? Or are you going to follow the target and sort of play it one by one?

That's a great question, Mike. We do see, as we look at the industry, the relative proportion of small molecules versus antibodies now. It is very significant, it is pretty balanced across the many companies' pipelines. And that's why moving into biologics was so important to us. It effectively doubles our targeted universe that we can now address with our technologies. So as we start up right now, we are looking to really show the first sort of proof-of-concept experiments, the first of early pipeline developments in 2023.

And so we -- because we bring the automation laboratory online for our biologics platform. And we do expect actually it to be an important part of both our internal pipeline, but also an important part of our partnership pipeline actually. And that's something which we would -- I think you should expect to see in 2023, how actually we develop the biologics platform in both of those directions. And we're excited by that. It increases the end-to-end platform's ability. I think what's really exciting about this as well, Mike, is that we've already demonstrated that our precision medicine platform works in biologics. This is already part of the deal we have with Sanofi and is absolutely ongoing work happening right now.

So what's important of our view, I think, compared to maybe ever sort of biologics sort of discovery organizations that are out there, is that for us now, it becomes a new modality that plugs into our end-to-end platform that can benefit from the target discovery engine that we're also using on the Sanofi platform. And then the molecules that we're developing could also then fit into our precision medicine strategy that we described today.

So what I would say is we do expect this to be a major part of our pipeline going forward. But right now, I wouldn't give guidance on the exact balance in small molecules and biologics. But what I would say is that as you've seen the field develop in the rest of the industry, now that we have a wider range of targets we can go after, the biology will dictate actually, which is the best modality to then go for.

Right. Makes sense. And then just a quick follow-up on that. I mean, is there anything we should take in terms of the improvements that the platform can provide for biologics drug development? And by that, I mean you've characterized the value add for the small molecule side of things, should we -- when we think about biologics, should we think about it being sort of a similar benefit in terms of cost reduction, time reduction of getting a drug to [IND] phase?

Absolutely. That's why we are taking a generative design approach. And actually to give you a bit more color on the technology development that made us so excited about development platform, as you want to bring Garry Pairaudeau, CTO into this conversation. Garry?

Yes. Thanks, Andrew. Hi, Mike. Yes, I mean we -- as Andrew said, we're super excited about this. And I think that the benefits are going to come in 2 manifolds really. So biology exploit design allows us to target specific epitopes allows us to target proteins that aren't accessible through traditional methods. And of course, we're also going to get a speed benefit and a quality benefit by being able to design biologic molecules that are intrinsically more developable and more suitable for the treatment of patients, better humanization, et cetera.

If I can squeeze in one -- one quick one for Ben. I appreciate the color on the cash burn and cash balance. Any high-level guide point post what you think about for 2023 as you've got candidates progressing into the clinic in terms of how to think about R&D expense or cash burn?

Yes, we're not giving any formal guidance. What I'd say is -- there probably will be a slight uptick, but not something dramatic from where we are this year.

Your next question is from the line of Chris Shibutani with Goldman Sachs.

This is Charlie on for Chris. I just had a couple of questions on the '546 program as we're moving into the clinic into the new trial there with this ABS patient signature that you've identified. I'm just wondering, how does that work on a practical -- from a practical standpoint in the clinic? How easy is it to identify patients by this signature? And as we look down the road in terms of a potential future label, how are you thinking about what that label will look like in the context of a signature like this? And how would that look in terms of if you were trying to go after a tumor-agnostic approachable label. Just wondering if you could provide any color there.

And then additionally, on '546, I was just wondering, as we're looking down the road towards potential checkpoint inhibitor combinations, how you're thinking about which combination partners you could potentially pursue and whether a formal partnership based around one of those combinations if possible.

Thanks for your question. Good to speak to you again. I'm going to actually hand over this question to Mike Krams, our Chief Content Medical Officer, to guide you through how we're thinking about using our biomarker signatures in our clinical trial planet. Mike?

Great question. So ultimately, our objective is to be a matchmaker between the problem of the patient and the solution that we provide. To achieve this, it's important that we identify biomarker signatures that enable us to match the right treatment to the right patients. As we start our Phase 1/2 study, and you will hear more about this soon, and we are going to further qualify and build confidence in the biomarker signature. And the design is set up to tell us at what level we have sufficient confidence in the biomarker signature to provide to us a decision-making quality and when to start to in which patients.

So our first Phase 1/2 study will learn about the compound, but it will equally learn about the biomarker signature. And the minute we have sufficient confidence and it being deployed as a decision-making tool will switch and deploy it accordingly. I really think that the second question is absolutely critical. And it speaks to our philosophy.

We developed drugs in a patient-centric fashion. So this is not about just developing a particular compound. But the question is, what is this compound contributing to the overall solution to the patient. In the case of an A2A receptor antagonist, we clearly need to think about combinations where checkpoint inhibitors. And the entire program in Phase 1/2 is indeed set up to teach us on how to think about the add-on strategy of A2A receptor antagonist on 2 checkpoint inhibitors and more on that soon.

Great. If I could just squeeze one more quick one in just for my own curiosity. Now that -- we're seeing the biologics part of the platform ramping up as well. I'm wondering if you see -- if you can envision the possibility that you might see kind of a combination between the small molecule and biologics programs where you might be developing ADCs for a very specific patient population?

That's a really interesting question, Charlie. Actually, the flexibility we have within our platform actually gives us a wide range of possibilities of what we could go for. This is what's exciting actually by having these different sort of design engines at our disposal is that actually such a possibility of design in ADCs is now actually within our wheelhouse. And it's something I try to know what actually many of our sort of targeted indication teams have been thinking about the possibility now of whatever modalities and combination of modalities actually to be put together with the 2 design engines we now have onboard.

Your next question is from the line of Peter Lawson with Barclays.

This is (inaudible) on for Peter. Congrats on the quarter. I just wanted to know if there's -- you could add some color to the changing sentiment around AI from pharma companies and what are you seeing at the -- are certain pharma companies that are more likely to do collaborations and what would convert the slower ones.

Great question, actually. In fact, we signed an increased interest actually in the use of AI within big pharma companies. A couple of years ago, it still felt we went into the half minds in how we are moving forward in totally large companies. What we're finding now actually that I believe that battle actually has been won. I believe actually people now see that this is the way forward that actually drugs are going to be designed and developed using AI. I think actually a proof points that we've already brought to the table showing that we already have molecules moving forward, showing the kinds of patient selection strategies that we talked about today and the results we've seen in things like first AI-driven sort of clinical trials of EXALT-1 and now coming together into a crescendo of showing potentially the direction of trial risk comes.

What I would say is that we continue actually to have deep discussions with any companies. We hope to be unveiling in the next few months as well sort of further collaborations from the MD Anderson collaboration by yesterday, to sort of deeper collaborations with pharma companies going forward.

But also what's important as well is that once we start collaborate to a company, also the continued deepening interest. Our deal will not be started off actually with one collaboration working on biospecific small molecules and from that one sort of in-licensed lead molecule, then led to a $5.2 billion deal.

Our work with BMS, already, we've licensed in the first molecule, which we'll be hearing about hopefully in the next few months and more to convert. And that then expanded the deal than originally for 3 projects now to 8 projects. All of which now are ongoing. So it gives us real confidence as well that we see our major partners when they start working for us really come back and more than double down on their work.

To give you a bit more insight into it, particularly on how those relationships are developing. I just want to get Dave Hallett, our Chief Operating Officer, actually, to give some of his thoughts. Dave?

Thank you, Andrew. Yes, I think you said the bulk of it already. I think the -- what we're seeing is using BMS and Sanofi have obviously as 2 significant examples are organizations who engaged with us many, many years ago in terms of a first generation of a collaboration of both Sanofi and BMS. And then based on the output and the productivity that we're able to kind of show to those organizations then came back to the table to obviously to do not only expanded deals but also to gain access to the expanding capabilities of our organizations.

And I think that's the journey we will continue to pursue is that to find kind of larger pharma organizations that are open to kind of utilizing the way that we're trying to reengineer the entire drug discovery and development process and to kind of join us on our journey.

So I think we will see more of those collaborations, but I think as we've reiterated on a number of occasions is that I think -- our preference is to go very, very deep with a subset of major partners. There are very good operational reasons for that running. That's the direction I think you'll see from a business development perspective over the coming years.

Andrew, just one point I wanted to add on, this is Ben. What we've really seen is the commitment from the top makes a lot of difference. I think if you look at all of the organizations across the industry, they are all convinced that technology is going to change the way that we do drug discovery, they're all doing different experiments. But if you look at the ones that do the larger transactions, really the sort of things that we focus on because to Dave's point, we like to go deeper. We like to do larger deals. Those are really going to be deals where they've got the executive buy-in and they're making it a priority in the strategy.

So that's one of the biggest items that we look for when we're evaluating our partners. And just one last word from Mike Krams as well as also I'd like to add some color to his question.

Yes, it's really just like -- to highlight the experience that I've had coming from big pharma and joining Exscientia recently, what strictly has been the interest and ability to connect the dots and to integrate AI expertise and experience in the discovery platform. AI expertise and experience in the translational science platform and now the ambition of applying the I in a clinical development platform. What I haven't seen before is the tight integration and the connecting of the dots in a very strategic fashions in an end-to-end ambition, which we at Exscientia here have written on our business cards in a patient-centric manner.

Your next question is from the line of Vikram Purohit with Morgan Stanley...

This is (inaudible) on for Vikram. We have 2 quick questions. So for the recently announced collaboration with MD Anderson, could you discuss the terms of the agreement? And what the focus areas in oncology are. Additionally, your release notes that your Sanofi collaboration has progressed an additional target in oncology. Could you speak about the review process this target went through and what the next steps are? And could you discuss how specifically Exscientia's platform was used to identify this target?

Great question. And actually, the best (inaudible) actually Dave. Dave Hallett?

Thank you, Andrew. So let's start with the question on MD Anderson. So as we announced yesterday, the -- the collaboration with that group is really predicated on a couple of major concepts.

One is leveraging (inaudible) small molecule design capability but also its translational platform. But then marrying that with the significant kind of discovery and clinical scale research that is conducted at MD Anderson. So when we sat down over the course of summer to kind of put this collaboration together, is that -- what I particularly liked about the relationship is -- as we're looking through potential targets that we could discuss and evaluate that would come into the collaboration. We were -- we had kind of principally investigators from MD Anderson kind of in those calls.

These are physicians who are interested in the biological pathways that we're contemplating. So I think it's kind of highlighted in the press release is that we have a more than interested kind of clinical set of investigators within MD Anderson. So should we be successful with this collaboration I think we've already got a potential kind of path into at least these clinical trials.

I'm not -- kind of liberty to go into the kind of financial terms of the collaboration, but this will be a joint venture between both parties. I'm personally kind of very excited to be working with MD Anderson and with a very old former colleague of mine of Philip Jones, who I once worked with kind of over 20 years ago at Merck.

In terms of Sanofi, so I'll talk -- I mean rather than answer that third target, in particular, let me give you a sense of the broad way that we're kind of approaching that collaboration and kind of the -- what is the additional value that Exscientia brings to target identification. So the kind of things that we're doing is that using Exscientia our central biology platform is that we're doing significant disease area landscaping within the areas of interest for Sanofi, so in subsets, specific kind of landscapes around kind of oncology indications, specific landscapes around areas of kind of inflammation and immunology.

That post -- what we've been able to integrate there is not only the kind of corpus of -- kind of public data and applying the deep learning methods the reside they're in is that we can also integrate and have integrated with that proprietary data sets like GWAS data sets, for example, that's Sanofi have brought to the table. And so by integrating, if you like, the know-how from the 2 organizations on top of the -- kind of AI capabilities, that's allowed us to identify kind of significant -- kind of short list of kind of potential targets to explore in more detail, including experimentally.

But also -- but part of that experimental validation of kind of those -- kind of targets is that we can -- as well as doing kind of functional genomics, et cetera, and working cell lines, one of the beauties of the collaboration on behalf of Sanofi is that we can do further validation in primary human tissue settings on our precision medicine platform. So not only validating the target from a biological pathway perspective, but from the very start of the program, highlighting the potential patients we might want to go after in the clinic.

And Mike, did you want to add something as well to that? Sorry, Mike, you want to... Sorry, Mike...

I think Dave you summarized (inaudible) really not much to add.

Okay.

There are no further questions at this time. I will now turn the call back to Andrew Hopkins.

Great. Thank you very much, operator. Thank you to everyone on the call today. Thank you for your continued support to Exscientia. All commitments and into 2023, we look forward to advancing multiple programs going forward, including '546 and '617. But this work is just the beginning for us, and we remain excited by the potential of our AI and deep learning platforms that's the whole for future of drug development. Thank you again for joining us today. And operator, you may now disconnect.

Ladies and gentlemen, thank you for participating. This concludes today's conference call. You may now disconnect.