Scroll Top

T Cell Engagers: The New Cancer Drug?

The Harry Glorikian Show 

Leonard Wossnig, LabGenius 

For January 16, 2024 

Final Transcript 

Harry Glorikian: Hello. Welcome to The Harry Glorikian Show, where we dive into the tech-driven future of healthcare.  

One of the most amazing successes in the battle against cancer over the last two decades has been the introduction of antibody drugs that harness the body’s own immune system to kill tumor cells.   

Finding those drugs may sound like a biology problem rather than a machine learning or a big-data problem.  

But actually, these days, it’s both.  

My guest today is Leonard Wossnig, who’s the chief technology officer for a UK company called LabGenius.  

The company uses a combination of synthetic biology, high-throughput assays, and machine learning to hunt for new drugs within a subclasss of antibody medicines called T cell engagers. 

What are T cell engagers?

Loosely speaking, T cell engagers are a new antibody medicine that can grab tumor cells with one end and then grab tumor-killing T cells from the bloodstream with the other end. Leonard Wossnig says the key to the whole thing is having the best data possible.  

Meaning, data about their candidate T cell engagers and how specifically they bind to their targets in the lab assays.  

In an ideal world a T cell engager would ONLY bind to cancer cells.   

In reality, many candidate antibody drugs also have off-target effects, meaning they also end up binding to and killing some healthy cells.  

LabGenius has built an automated platform called EVA that runs experiment after experiment to find T cell engagers with just the right binding strength.   

Basically it uses the data from each experimental round to compute the optimal design for the next candidate molecule in the next round.  

And in this way, the company hopes to identify T cell engagers that have so few off-target effects that they can confidently take the antibodies into clinical trials.   

Wossnig has a background in physics and computer science, so it’s no surprise that he’d be coming at this whole problem from a data science perspective.   

In the interview, you’ll hear how he keeps coming back to the primacy of data. 

One of the big takeaways is that companies that want to use AI to speed up drug discovery need the biggest, cleanest, and most consistent data sets possible.  

So here’s my conversation. 

Harry Glorikian: Leonard, welcome to the show. It’s great to have you on again. It’s been a while since we talked last. You had an exit, and, uh, you know, you’ve joined this company called LabGenius, and I just wanted to have you on the show and talk about it, because if I’ve done my reading correctly, LabGenius is in the business of using lab automation and machine learning to find more effective antibody medicines. But I was looking at that and I’m like, there’s a heck of a lot of detail embedded in everything I just said. Um, I know I’m using simple words, but there’s a lot of detail there. So I’m hoping that, uh, just so the listeners understand the context, and the huge stakes. Uh, I wanted to ask if you could start by talking about the basics, and then we can sort of maybe go step by step into the actual work LabGenius is doing. So, I don’t know, maybe we we start with why antibody therapies? You know, how do antibody therapies work, what kind of diseases do they treat.  

Leonard Wossnig: Okay. Well, it’s a very broad question, but yeah, therapeutic antibodies have already transformed like countless lives, right? Today. I mean, they are among some of the biggest selling drugs in multiple cancers and other indications. Um, but we also definitely see and that’s not for not true only for antibodies, but also for many other drug classes , there’s a decrease in R&D productivity. And I think one of the, you know, one of the potential reasons is that as our understanding of diseases has matured over the last 10, 20 years, we also are moving to more complex therapeutic agents, meaning these the type of antibodies we’re dealing with is becoming increasingly more complex. And I think one of the big challenges that we face when we want to optimize these molecules to be ultimately drugs for patients is that the process of optimizing, meaning finding these really these new next generation antibody therapeutics, is becoming more and more complex and hence requires new tools. And that’s that’s what LabGenius is focusing on.   

Leonard Wossnig: Maybe a few words, as you said, about antibodies. Antibodies are inherent to our immune system. You have all multiple well, uh, incredible large number of antibodies in our body. In drug discovery, we repurpose those. We, we build antibodies specifically to do a specific, to have a specific function. In LabGenius’s cases, we focus on a subclass of antibodies called T cell engagers. At the moment, with our pipeline, a T cell engager, without going into too much detail of the biology, but in a nutshell, a T cell engager is a molecule which on the one hand tries to grab—and I’m using this as a, you know, representation—and tries to grab a tumor cell, for example, on the other hand grabs a T cell. And when those come close enough in close proximity, then the T cell engager is activated and it releases specific chemicals or molecules, which then destroy the tumor cell targets in a targeted manner. And that’s how T cell engagers work. Does this broadly answer the question?  

Harry Glorikian: Yeah, I think I was reading somewhere that 30% of the industry is focused on antibody therapies now, and I think that could grow if we got better at discovering new antibody based medicines. But. I mean, why is it so hard? What are the current limitations? And why is it so slow and expensive, considering all this technology we have right now?  

Leonard Wossnig: Maybe let me talk first about why it’s so hard and I think, that’s, I’ll probably try to focus the answer specifically on what we are working on as well to kind of be more precise in answering it. Um, of course, when we look at drug discovery in general, the question why is it so hard is because there’s a lot of different properties for these molecules that we need in order to for a molecule to make a therapeutic agent. So just to give you an example, we need to have a molecule to be potent enough. So it needs to really have it a certain concentration, kill enough of the cancer cell. We need to, at the same time, we require this molecule to be selective, meaning that, you know, it doesn’t destroy any other cells, particularly healthy cells in the body. So activity is a big issue because that’s related to safety in the patient and side effects. On the other hand for, and that’s more around production and so on. Every medicine needs to be produced at scale. And also of course, if you want to administer it to patients, we want to make sure, um, it’s easy to inject it, for example, in a subcutaneous way.  

Leonard Wossnig: So besides the functional properties, potency, selectivity, we also need multiple properties of the molecule to make it manufacturable and stable, so we can transport it, we can administer it, etc. as well. So there’s a quite a large range of properties we desire in these molecules in order to really make it a valid drug in the end. And, and I think, you know, given the amounts, the incredibly large space of potential molecules which could satisfy these properties, the search that we currently using is extremely inefficient. That’s because most of the searches are still done by hand. And that takes, you know, for even a small handful of antibodies, several weeks for a human in the lab to test and make these antibodies. And if you want to search a space of, you know, hundreds of thousands or even millions of potential molecules, of course, doing five at a time in a couple of weeks will never enable you to allow to test them all and find the one which satisfies all the different properties that we need for a drug.  

Harry Glorikian: Yeah, I remember doing things by hand back in the day. I don’t think that’s efficient anymore, but I’m dating myself for a long time ago. But if I was reading correctly about LabGenius, you guys are actually hyper focused on one specific problem with antibody therapies, and I think you guys call it off-target effects. Or in the case of cancer immunotherapy, off-tumor effects. Can you describe. It’s sort of off target effects in basic biological terms.  

Leonard Wossnig: Yeah, absolutely. So typically in many cancers, if you think about it, a cancer is just a number of cells, right. And each cell has on the surface a range of receptors. So these are just large molecules. And typically what we have is like a lot of antibodies target these receptors meaning they bind to the receptor. And then based on the binding of the receptor they recognize whether it’s a cancer cell or not. So that’s all good and well. The challenge arises if the same receptor is on the healthy cell as well present as well as on the cancer cell. And if the same receptor is in fact present as is for many cancers, the therapeutic agent or the antibody specifically will destroy both the healthy and the tumor tissue. So we are focusing on on molecules which, you know, are built in a way such that they recognize the difference between a healthy and a tumor cell. And they do this by selectively binding only to the tumor cell. And that’s what we’re trying to focus on. Because what we’ve seen in a lot of particularly T cell engagers in the clinic is that they’re very potent. They they do bind the right receptors. But they also unfortunately then have a side effects. And these come typically from on-target, s  o binding to the right receptor but off tumor engagements.  

Harry Glorikian: So this is a Goldilocks problem right. The porridge is a little too hot or a little too cold. And you’re trying to find the right one in the middle. That’s just right. If I remember the story correctly.  

Leonard Wossnig: Correct. Correct. Yeah.   

Harry Glorikian: What are the conventional methods for solving this problem? I mean, what are the—I’m trying to figure out, what are you guys doing that is different than how we’ve done it in the past.  

Leonard Wossnig: Yeah, I mean, conventionally, maybe it helps to explain kind of how we go about identifying these antibodies and what type of antibodies we’re building. So LabGenius works with something called VHH’s, which are also known as nanobodies. These are essentially small mini antibodies which are quite stable and have a number of other benefits. Now we take these individual antibodies, these small nanobodies which bind, for example, the receptor or, or something else. And then we combine these individual binders into more complex multi-specific and multivalent antibodies, meaning these form—think of each of these binders as a little hat, and we string these little hats or pearls together on a chain so that they can bind to, essentially, for example, multiple receptors at once. And by targeting the right number of receptors at the same time, we we’re using a so-called avidity, avidity driven approach. Now what does this mean. So imagine you have say my fingers. You see them, right. These could be the number of receptors on a tumor cell. Now, if I take like two of these off, then say in a healthy cell we only have two of the receptors. Now, what you’re trying to do is we find, we try to define a counterpart. So an antibody which has the right number of hats of these, you know, these binders, these cages, so that only if all four receptors here are present, now I think the camera is kind of preventing me from doing this correctly, but we tune the interaction strength of the individual heads of the individual antibodies and, and the number of these antibodies which are in there in one of the chains, so that only if all if the receptor number is high enough, only then they interact and bind. And of course, if you don’t have the four present but say only two of these, then the interaction wouldn’t be strong enough so the antibody wouldn’t bind. So meaning there’s a lot of different, you know, things we need to test. For example, how close together do these individual heads or antibodies, these nanobodies need to be, how flexible or rigid are they. Do they need to tie together.  What kind of binding strength to do these antibodies need to have. And by testing all these combinations we ultimately find the right one. Now, how humans are doing this is by typically trying to think, okay, based on the early data that they received, they’re trying to manually change the linker length, change the affinities or the binding strength, and then iterate through this process quite slowly and, and in a manual fashion. And that’s, of course, quite slow and time consuming, as we discussed earlier. And, and hence, um, not very successful, particularly if you want to explore very large design spaces. We use a combination of machine learning, robotic automation, um, and and robotic automation to really accelerate this process, to search and make it much more efficient so that we can, you know, test out all these combinations in a manner,  at a speed which is unprecedented and hence allows us to identify, you know, the right number of these heads, the right strength of binding and the right kind of, um, distances, etc., and linker rigidity in order to find the right molecules.  

Harry Glorikian: You call that Eva, which is, uh, what is it, Evolution Engine, if I remember correctly, what it was called.  

Leonard Wossnig: Correct. Yeah, yeah.  

Harry Glorikian: So how do each piece of that evolution engine really work and, and what key technologies are coming into play at each step?  

Leonard Wossnig: Yeah, so, so, the key technology really, which is underlying this entire technology stack that we’re using is something called active learning. So active learning, or also sometimes known as optimal experimental design, is an approach where you use machine learning in an iterative manner to always predict what’s kind of the best set of experiments to run at a given stage. Then you learn from these experiments. So you update your models, your machine learning models, and then you predict the next one. And so you kind of progressively accumulate the data in an optimal fashion to ultimately drive you to the desired result. In our instance, we use an active learning approach called multi-objective Bayesian optimization. And this means that we take all the different properties that the molecules, that we desire these molecules to have, and then we optimize or co-optimize, in fact, all the properties at once. And we try to search the design space with this active learning approach, so that we find the highest performing molecules in this incredibly large space of potential molecules as efficiently as possible. So I can explain how this works in a bit more.  

Harry Glorikian: I’m trying to imagine all of this happening at one time efficiently.  

Leonard Wossnig: Yeah, so, so, how can you imagine this? So I so we follow kind of this loop, essentially. So we start with the design of the molecules. That’s all done in silico, computationally. We then produce these molecules. In our case that means cloning, expression, purification, and then ultimately running those, like, testing these molecules in functional cell based and disease relevant assays. Then we collect the data, we upload the data, and we start the whole process again. So when we collect data, we as mentioned previously, we collect data for all the different properties. That means the potency, the efficacy of the molecules, the selectivity, as well as multiple developability criteria such as expression yields. So how much, how easy it is to produce the molecule, thermal stability and kinds of other developability criteria which are essential in, in the design of these antibodies. Then once we collect the data, we upload this all to our cloud- based platform and then use the data directly to retrain these machine learning models. Then based on the on these updated machine learning models, we make new predictions, meaning we design new candidates and then repeat this whole process. And that’s kind of how we typically go in a program. It takes around 4 or 5 cycles to deliver high performing molecules, which satisfy all the criteria we need for these.  

Harry Glorikian: So how long does that usually take?  

Leonard Wossnig: So currently I think that’s, honestly, that’s world-leading. We require less than six weeks for around 768 molecules. And these are to be like, that’s, that’s quite important. These are not simple binding assays. You know, that’s what we use. But these are really disease relevant and cell based functional assays.  

Harry Glorikian: So 6 to 7 weeks from start to finish.  

Leonard Wossnig: Uh, less than six weeks.  

Harry Glorikian: Less than six weeks. Okay.   

[musical interlude]  

Harry Glorikian: Let’s pause the conversation for a minute to talk about one small but important thing you can do, to help keep the podcast going. And that’s leave a rating and a review for the show on your favorite podcast player.  

On Apple Podcasts, all you have to do is open the app on your smartphone, search for The Harry Glorikian Show, and scroll down to the Ratings & Reviews section.   

Tap the stars to rate the show, and then tap the link that says Write a Review to leave your comments.   

On Spotify, the process is similar. Search for The Harry Glorikian Show, click on the three dots, then click “Rate Show.”  

It’ll only take a minute, but you’ll be doing us a big favor, since positive ratings and reviews help other listeners discover the show.  

Thanks. And now, back to the show.  

[musical interlude]  

Harry Glorikian: Let’s step back here for a moment. I’m going to I’d like to sort of dig in a little bit into your personal background. How did you how you came to LabGenius, right? I know you joined in 2022. Uh, before that, you were with a company called Odyssey Therapeutics, uh, which had acquired your startup Rahko, if I’m pronouncing it correctly. Tell everybody about your background, your training. What was Rahko all about? I mean, I understand, um, or I know you were developing a quantum machine learning model for drug discovery, and then maybe, what inspired you to leave Odyssey and join LabGenius?  

Leonard Wossnig: Okay. Yeah. Many, many questions. Let me start with the with the background. So initially I saw very early on, I come from a physics background. I moved then though during my PhD into computer science. So I was a Google PhD fellow at UCL in London, um, focusing on a lot of very mathematical topics, including statistical learning theory and then quantum machine learning. And during my PhD, that was to a large extent due to my PhD advisors. We founded the company. So me and a couple of other PhD students co-founded this company. Um, the focus on Rahku was really, you know, these quantum machine learning models. And that specifically, these models are, we’re trying to make good predictions with minimal amounts of data. So what we’re really trying to achieve there is to use as much prior knowledge, particularly physical knowledge, you know, about molecules and so on. Um, and encapsulate this knowledge into these models to make sure they can learn as efficiently as possible. And of course, the big need of this is that in most drug discovery campaigns as well as material science, often you don’t have the amount of data which is needed, you know, compared to, um, large language models today where you can basically train your models on entirety of the internet.  

Leonard Wossnig: So in drug discovery and in many of what material science applications, similarly, you really need to rely on very small data sets, in contrast. And in Rahku we really tried to solve the problem of how can we make as good predictions in practical context as possible using the minimal amount of data? That was t he background. Um, of course. Then I moved into drug discovery. The company got ultimately acquired by Odyssey Therapeutics, as you mentioned. I have to also add, there was the like, Rahku beyond the quantum machine learning. We did also do small molecule generative design as a another thing and and Odyssey is obviously a pioneer for all of these methods and is still doing very well today. You might have seen the recent announcement of another, um, large fundraise. So I think Odyssey has by now raised nearly half a billion. Um, and it’s, I think, doing quite well as well.  

Harry Glorikian: So then that begs the question of. Why leave Odyssey to come to LabGenius?  

Leonard Wossnig: Yeah. It’s a good question. So I think I think the what, what what LabGenius has and that’s I think quite unique is, LabGenius is really creating this, this closed loop optimization. And I think there’s very few companies which have developed a full technology stack where end to end, everything is focused on running these loops, these cycles where you design, build, test and learn. And I think, um, biologics are a beautiful application and specifically the way we have done this. Because with biologics you can iterate, you know, create data and learn from this data at a pace which is just, like, unimaginable in many other kinds of companies I think today. So I think LabGenius is probably one of the, if not the only company which can complete the cycles at a speed, really. Um, you know, and at this speed, meaning less than six weeks and producing more than 700 candidates. And then in these functional cell based assays.  

Harry Glorikian: How do you know the platform is working and that the investment is paying off? Um, I mean, I know you’ve had a research collaboration with Sanofi. I don’t know if you can talk about what you’ve learned from that work or how it bears out your, your, your central hypothesis or proposition about antibody drug discovery.  

Leonard Wossnig: Sure. So, I mean, we have multiple indicators that the approach is working. Maybe to say this ahead of time. Um, the Sanofi collaboration, uh, collaborations, is just one of those. And maybe I can share a bit of background for internal case studies, but also about the Sanofi collaboration, to kind of highlight why I think this, this approach is already showing very strong indications that it’s working quite well. Um, so in the Sanofi collaboration, we essentially co optimize different properties, um, which you know, is not a um is not an easy task in many. And that’s what, you know, everybody who’s working, whether in biologics or in, in small molecule drug discovery is very aware of there’s typically this whack a mole challenge, right? You optimize one property and the other one is then declining. And I mean, everybody who’s worked in a drug discovery program would just smile as you just did, right?  

Harry Glorikian: It’s like it’s like whack a mole, right?  

Leonard Wossnig: Exactly.  

Harry Glorikian: The problem moves.  

Leonard Wossnig: So and I think that’s that’s something obviously Sanofi was, was also very aware of. And so they wanted to test our platform and how it worked. And I think what we’ve beautifully demonstrated in that collaboration, I think the um, that’s public knowledge as well, is that we can, in fact, use machine learning in this approach to co optimize multiple properties at once, which is, I think for humans, a non-trivial task. So beyond this, um, we have also run internally a case study which we have published in multiple conferences already as well. And this case study was specifically a T cell engager targeting Her2, which, you know, very common breast cancer target, for example, and CD3, which is a receptor on the T cell. So in this instance we designed, we tried to design a T-cell engager which is highly selective. And this again comes down to the question, you know, in T-cell engagers, many of the markets have have off-target,  on-tumor, off-target effects. And we try to demonstrate that our platform can really meaningfully overcome those, those limitations by designing molecules which are highly selective.  

Leonard Wossnig: We use in this instance a benchmark molecule, which I think, um, currently in phase 1 or 2 from Genentech called Nimotuzumab. So that’s, that’s, you know, a molecule which I think is close to clinical or in the clinic today. And we then run our platform across five cycles. And, you know, we’ve beautifully seen with each cycle a drastic improvement in the performance of these molecules. And ultimately after five cycles, the, um, the top performing molecules were consistently outperforming the clinical benchmark. That’s all nice and well, but then we took this high performance, and tested those in T cell engager killing assays. So these are really indicative of how is your T cell engager now killing the cancer cells. And in these T cell engager mediated killing assays, we demonstrated that the whole panel of molecules we identified had in fact, uh, more than 10,000 fold selectivity, which is, you know, at least 400 fold greater than the clinical benchmark molecule, while still reaching 100%, killing and having a less than, uh, kind of think less than ten picomolar potency, which is still quite significant.  

Leonard Wossnig: So I think in a nutshell, there’s now based on the multiple proof points we’ve seen, I think this approach works. I think it allows us to design T-cell engagers, but also other multi-specific antibodies, which are really with an unprecedented selectivity. And I think going forward, we are just applying this now to our internal pipeline to hopefully come up with therapeutics, with therapeutic antibodies to help patients in in the future.  

Leonard Wossnig: And I’ll admit, I haven’t had time to read it because I just saw it was you just published, a paper in, I think it just came out Arxiv. Can you talk about some of the ideas in that paper? Because I wanted to. I just remember something along the lines of, I took away from it, it was something along the lines of AI and biotech is sort of window dressing, unless the company also has the technology to generate the data required for machine learning. So I wanted to sort of yeah, get your get your thoughts on on maybe talking about that paper and your, your position based on your experience.  

Leonard Wossnig: Yeah. So I mean, I think I’ve written a lot recently and the last year about how important it is for companies to generate high quality data. I think, um, there’s a lot of talk, obviously, currently, about generative AI and, and entirely novel approaches to designing molecules. But I think the reality is, if we look out in, you know, what’s out there today in terms of what data is out there really in the public domain to train these models, it’s mostly binding data. And for a lot of next generation therapeutics where we have new modes of actions, we have new modalities, we have new biology in general. There is, of course, no data available to train machine learning models on. And for this reason, I think, um, one example is T cell engagers, right where, um, functional data just barely exists in the public domain. And hence no machine learning model will be able, off the shelf or de novo or with zero shot design, to predict the functional readout of a T cell engager. And I think for this reason, it’s quite important for companies to build really solid data stacks. So the ability to generate data at the lowest level, but then also collect the data, store the data in the right way, process it in a manner so that the data is consistent. And ultimately only when all these layers are correct, then they can make use of machine learning methods to predict essentially how these molecules behave ultimately and hopefully use these predictions to find better ones. But without the proper tech stack and without the ability to really, yeah, do everything end to end, um, in, in the right manner, I think companies won’t be able to deliver meaningful value in the pipeline using machine learning.  

Harry Glorikian: Yeah. I mean, it’s, you know, just, you know, investing in the space, the number of companies that don’t have all the right people, all the right team members, you know, to to pull it off. Right. Because it’s the team, it’s the technology, it’s the data. Right. And if you don’t have all the pieces, it’s sort of you can see how they’re not going to be successful. Even when you have all those people, all that stuff, it doesn’t mean you’re going to be successful. It just means you’ve got the right, uh, ingredients to make, uh, the right dish, right to that you can then take forward.   

Harry Glorikian: So what does success look like for LabGenius? I mean, wat do you hope to be in five years? Is it partnerships? Is it internal drug development programs? Um, you know, is there. Another company that you’re trying to that you admire or you want to emulate in the tech bio space, right? Um, in other words, if you said. We want to grow up and look like X in five years. Or you know, what would X be? Or do you want to just be your own? Do you want to be the X that everybody else emulates? Right. So what is success look like?  

Leonard Wossnig: Yeah. Uh, it’s a good question. I mean, I think because we are a biotech success is is always quite, quite obvious. Right. And the thing which matters most for any biotechnology company is to, to bring, to find the right drug for the right patient and ultimately bring it to them and make sure it works. So we are currently mainly focusing on our internal pipeline. So we are progressing a range of different multispecific antibodies, um, with a big focus on T cell engagers at the moment. And success for us, of course, looks, um, like a positive readout in the clinical trials at some point. We at the moment are a preclinical company. So hopefully answer is [to be at] the clinical stages in the next two years or so. But obviously based on this, I would define success for us as becoming a clinical stage company and ultimately, um, bringing a drug to patients, which hopefully is potent, uh, efficacious and safe.  

Leonard Wossnig: I know that, you know, after talking to you before and talking to you now, there’s a lot of sort of big ideas, um, that you can talk about, but I sort of wanted to ask this big catchall question, right. Which to make sure like we haven’t missed anything important, you know, if you were interviewing yourself. And you really wanted to get to the core of what motivates you and LabGenius. What what questions have I not asked you that I should be asking you? And how would you answer?  

Leonard Wossnig: The one thing maybe just just worth highlighting, I think, you know, there’s currently a lot of talk about generative AI and machine learning, etc. I think, um, we’re still at the very beginning of this, these exciting kind of areas of development. I think there’s a lot of things to explore. And I think particularly as we are learning now, generation of the right type of data in sufficient quantities is key for machine learning to succeed. So I think maybe going forward, I’m very optimistic about, you know, what’s this area of broadly speaking, AI drug discovery is going to bring in the next couple of years because I think now we are just at the, you know, the almost a turning point where more and more people realize we need to create the right data sets today to make sure we have the, you know, the better fruits in the next couple of years. And you see this from small biotech to large pharma, right, where people start collecting more and more data. I think what’s quite important is and that’s, again, one of the big reasons why I joined LabGenius, is collecting the right type of data, the right quantity and quality of data is, in my opinion, key for success of this field. But I think a lot of people are realizing this right now, and I think we’re going to see the impact that the data that we collect today will ultimately have in the next couple of years. So I think I’m quite excited about what this this field is going to bring. And I’m looking very much forward to see more of these, the results in the next couple of years as well.  

Harry Glorikian: Yeah. I mean, when I think about. I mean, first of all, when I think about just, you know, the AI, machine learning space, language models, etc., I say to people, to use a baseball analogy, you have nine innings. You’re at the top of the first. Like, you haven’t even started. You haven’t really started the game yet. Or you’ve barely started the game. And then when I think about now, our world adopting these approaches, having the data, etc., etc., I mean, I feel like we’re even earlier technically. Um, because a lot of those other areas have data available to them that they can consume. Right. We’re still at that point of, yeah, we may have some places where we have data consumed, but there’s a lot of that. We don’t have enough data or good quality data or in the right format to be able to consume.  

Leonard Wossnig: Correct, correct. And I think also and that’s something, you know, particularly the antibody space, I think, well, we put out another paper recently that was with colleagues at Moderna and Sanofi and AstraZeneca. And the paper highlighted how important it is to collect consistent data, and consistency is something you can really only control by optimizing every stage in the process. Right? And this this means from what type of data you collect to what controls you use, how you collect your data, is it manual or is it automated to really every bit in this tech stack? And I think, um, that’s going to be a quite important thing because the companies who are building, you know, this, this tech stack end to end, and with the involvement of all the different teams, right, not only a tech team, not only a scientific team, but all teams together, I think they will succeed in creating the right quality data. And I think that’s a key requirement to ultimately have success, because I think, you know, looking in, I mean, a lot of the public data that we have available today, the quality is quite poor as well. There’s duplicates which disagree with each other. There’s a lot of noise. I think creating consistent data will have a big impact on how well these methods will work as well.  

Harry Glorikian: Oh, totally. I mean, I can tell you that throughout my career, diagnostics, whatever, I mean, you can see people making small changes and then the results in the end, just something doesn’t add up. Right? And people don’t understand. It has to be done exactly the same way for you to get the same results.   

Harry Glorikian: I’m so glad we could make time and get you on the on the show and talk about this. I mean, I’m, uh. I’m super excited about what you guys are doing. Um, and I, you know, I want to keep up to date on the company and how things are going. Um, and it’s the holiday season, so I’m going to say happy holidays. Uh, and, um, uh, look forward to keeping in touch and, uh, hearing how things are going. 

Leonard Wossnig: That’s a really nice to talk and good to see you again, Harry. And yeah. Um, absolutely. Looking forward to speaking the next time again.  

Harry Glorikian: That’s it for this week’s episode.   

You can find a full transcript of this episode as well as the full archive of episodes of The Harry Glorikian Show and MoneyBall Medicine at our website. Just go to and click on the tab Podcasts.  

I’d like to thank our listeners for boosting The Harry Glorikian Show into the top two and a half percent of podcasts globally.  

To make sure you’ll never miss an episode, just find the show in your favorite podcast player and hit follow or subscribe.   

Don’t forget to leave us a rating and review.  

And we always love to hear from listeners on X, where you can find me at hglorikian.  

Thanks for listening, stay healthy, and be sure to tune in two weeks from now for our next interview.