Summary

On the latest episode of The Bioverge Podcast, we're thrilled to welcome David Li, Co-Founder and CEO of Meliora Therapeutics.

David sits down with Neil Littman to discuss why the mechanism-of-action for cancer therapies in development is poorly understood, how this leads to a high and costly clinical failure rate. Without an accurate understanding of how cancer drugs actually work, patients will receive the wrong treatments - contributing to the unacceptably-high failure rate of new therapies

Meliora is bringing the latest genomic, computational, and machine learning techniques to bear to change this paradigm. Melioro believes that by identifying the true mechanisms-of-action of chemical compounds allows more accurate, effective, and rapid drug development.

Listen today wherever you get your podcasts (links below):

Transcript

00:09
Danny Levine (Producer)
You're listening to the Bioverge Podcast with Neil Littman. 

00:29
Danny Levine (Producer) 
Neil, we've got David Li of Meliora Therapeutics on the show today for listeners not familiar with David. Who is he? 

00:37
Neil Littman (Host)
Yeah. So David is the CEO and co founder of Meliora Therapeutics. As you mentioned, Danny He was previously the chief business officer at Everest Detection, which was an early detection liquid biopsy startup that was founded by Brian Slingerland, who, for those who are familiar with him, he was the CEO. And founder of Stem Centrics, which was a company that was acquired it was an ADC company that was acquired by Abby for about $9 billion a bunch of years ago. David was also an early employee and led commercial development, commercial operations excuse me, at Benchling, which was a life scientist software platform company. And I think they're up to about a $6 billion valuation these days. So David led a team in sales, marketing BizOps, was instrumental in securing some of the company's first biopharma clients. David began his career in Ibanking. He was at Goldman Sachs and later moved to the investing side of KKR. 

01:36
Neil Littman (Host)
And I'm really excited to talk to David about what Meliora is doing because they really believe there's this confluence of factors that are contributing to a lack of understanding of the true mechanism of action, or MOA, for compounds being developed and brought forward into clinical studies in the oncology space. And so there is a very high failure rate for cancer clinical compounds, and I think it's about 97% of cancer therapies fail during clinical testing. And so David's hypothesis and what they're working on at Meliora is to better understand the MOA, to try to reduce that clinical failure rate. And so I'm really excited to talk to David about what they're building, what the platform entails, what are some of the inputs, what are the outputs, and how will that actually translate into increasing the probability of successfully developing cancer drugs? 

02:36
Danny Levine (Producer) 
We think about the elegance of biology and the ability to target the molecular drivers of cancer today. How much of a reality check is Meliora for the entire area of therapeutic development of cancer? 

02:50
Neil Littman (Host)
Well, it's a really good you know, I'm excited to talk to David about that. Obviously, we have made tremendous strides in developing targeted therapies for, right. There's no question. We have definitely moved into more of this realm of precision medicine. We're looking at targeted therapies that are really targeting the underlying molecular drivers of very specific cancers. And those drugs are highly effective for specific patient populations. Right. A lot of patients don't respond to those drugs for whatever reason. But there's no question that we've made tremendous progress over the last couple of decades with these targeted therapies. I think what Meliora is saying is, yes, that's true, but there's still a lot of progress that needs to be made to better understand and elucidate the MOA for a lot of these cancer therapies. And if we can better understand the mechanism of action that underpins some of these therapies, we should increase the probability of approving more drugs, right? 

03:53
Neil Littman (Host)
So if we have a better understanding of the MOA, we can more precisely target the correct cancer, and that should increase the probability of success. So that's my understanding of what they're doing. And so I really like their thesis. And of course, they built this sort of modern platform to be able to do this. So I'm excited to dive into what is that platform and how is that actually going to translate from their predictive platform to developing actual drugs. 

04:21
Danny Levine (Producer) 
Well, if you're all set, let's do it. 

04:23
Neil Littman (Host)
Danny, David, thanks for joining us today. I am incredibly excited to welcome you to the show. 

04:32
David Li (Guest)
Thanks, Neil. Really excited to be here and looking forward to the conversation. 

04:36
Neil Littman (Host)
So, David, today we are going to talk about the challenges inherent in developing novel cancer therapies. The company you co founded, Meliora Therapeutics, and your effort to leverage machine learning and recent computational advances in biology to develop the next generation of mechanism focused oncology drugs. There's a lot to unpack here, David, so let's jump right in. For the last 25 years or so, new cancer treatments have been defined as targeted therapies, falling within the realm of precision medicine. It turns out, though, that these terms might mistake reality. Let's start from the 30,000 foot view with the problem of developing novel cancer therapies and why it's so challenging. What is the success rate in clinical trials today for developing novel cancer therapies? And why do so many novel and seemingly promising drugs fail in the clinic, especially when so many of them have cured mice of cancer? 

05:32
David Li (Guest)
That's a great place to start, neil, there's been lots of ink being spilled about the increasing cost of drug discovery and how much is being spent per molecule, how slow it is. Amongst the kind of most important factors to these mounting challenges that we are facing in drug discovery is that approval rates are low. And in cancer drug discovery, it's particularly bad. Over 97% of cancer drugs are cancer drug candidates are failing in the clinic, and that's actually the highest failure rate of any indication. What we've seen over the last 1015 plus years is this kind of phenomenon where many different drug developers are throwing kind of the kitchen sink against the disease. Any agent that has anticancer effect is in mouse and in preclinical models are then being brought into the clinic to see what type of effect it might have in humans. I think there's a few contributing factors to this. 

06:28
David Li (Guest)
One is that obviously there's very high unmet need patients, everyday matters for the patient. And secondarily, we think that because of this kind of phenotype output we're looking for, which is really any agent that is killing cancer cells, there's often a glossing over of what is the proper mechanism of action that's actually happening with this particular agent. And so as long as a molecule is kind of killing the cancer cells at a rate that has the right window for safety, then that's good enough and let's move it into the clinic. And that, we think, is a really big missed opportunity, but also a really big factor for why approval rates are so low. If we don't understand the right mechanism of action, that is a very significant contributing factor for why ultimately many of these agents are unfortunately not being successful in the clinic. 

07:19
Neil Littman (Host)
Yeah, David, I think that's a really interesting view. And so, as we know, a lot of the targeted therapies these days are designed to target the underlying molecular drivers of a specific right, which on the surface sounds like a very elegant approach to targeting cancer and in many instances has worked well when that drug is efficacious in a given patient population. You talked a little bit about some of the challenges, but what do you think that we've gotten wrong? Does it boil down to a lack of understanding of the MOA before moving these drugs into the clinic? Or there are some other challenges? Or what's your view on where the failings have occurred? 

07:59
David Li (Guest)
Certainly there's a few different factors. And maybe I'll just give an example to start off this because I think it's pretty instructive. The first point you made is absolutely right. We actually have made tremendous progress in cancer care with targeted therapies. The example perhaps of Glivec is probably one of the best. It's one of the earliest. Targeted therapies is kind of hailed as a miracle drug for CML, a type of leukemia, and it was originally developed as a Bcrable tyrozine kinase inhibitor. Later on it was revealed it was also acting through a few other targets, including Kit and PDGFRa. And I think speaks to the fact that a lot of our current tools and tools we've been using for the last 10, 15, 20 years with regards to understanding mechanism are actually just proxies for understanding what this molecule is actually doing. One of the key standards being used is an understanding of binding. 

08:58
David Li (Guest)
For example, is this molecule binding to target X? And in our view, binding is a necessary but not sufficient threshold for determining what is the true mechanism of action of a molecule. There are several blind spots that can happen. For example, you might have a molecule that is really potently binding to target X. Now, if you didn't test for target Y or target Z, but that molecule may actually also have activity against those targets, and it may actually be that activity which is driving the anticancer profile of the molecule rather than its original interactions with target X. And one of the challenges with kind of the existing consortium of tools we have is that if you don't know a priori what to test for, often it's very difficult to determine what is the actual driving mechanism and activity that is truly giving this molecule its particular anticancer profile and to loop it all the way back to why this matters, then you're really trying to understand where to put that molecule in terms of indication and where to clinically deploy this molecule and thus give it a much better chance of success. 

10:13
David Li (Guest)
And so we think that if you don't really critically understand the mechanism and sometimes it's easy to get it wrong, that contributes significantly to, again, this kind of unacceptably low success rate of cancer therapies in the clinic. 

10:26
Neil Littman (Host)
So David, let's dive into that point a little more. Very specifically, your scientific co founder Jason Sheltzer co authored a study, I believe it was in late 2019, that suggests a drug's most potent target is not necessarily the source of its anticancer activity. So that's just what you were talking about, that it may have an activity that is not what we think is the MOA at the time. This was super interesting and it was a non obvious conclusion and it made waves in the scientific community. Could you talk about that specific study and what it found? 

11:05
David Li (Guest)
Absolutely. This study was the genesis of the scientific journey that Meliora started in that study. I'll just summarize very briefly what Jason and his lab was able to do. They essentially looked at a handful of cancer drug candidates, some of them pre clinical, some of them clinical stage. And he did CRISPR knockout of the purported targets of these molecules in cancer cell lines and he found a couple of pretty interesting results. One is that oftentimes those cancer cell lines with the targets removed didn't particularly care that happened and they just kept right on growing indicating that perhaps that these targets were not really critical for cancer growth and malignancy in the first place. So that's already interesting. Secondly, though, perhaps even more controversially, it's that when then the group used those cancer cell lines with the targets ablated via CRISPR and treated them with the original molecules, the anticancer agents, he found that many of those molecules still had anticancer activity indicating that it was not the original purported target. 

12:19
David Li (Guest)
That was the inhibition of that original purported target that was driving the anticancer effect. It was perhaps a different mechanism or a different set of mechanisms that was enabling these molecules to have, again, this anticancer profile. So it really speaks to there may be a whole set of molecules that are mischaracterized and actually mechanism is really difficult to get right because there's a lot of folks who've spent a lot of time and money on those molecules in order to develop them and again bring them forward into the clinic. So I think just another point that kind of speaks to why this is really critical for us to get right, find the right mechanism, put that molecule into the right patient population, develop the right appropriate biomarkers and thus dramatically improve our probability of success when we get to the clinic. 

13:09
Neil Littman (Host)
So David, is it fair to say that a fundamental misunderstanding about the mechanism of action of a lot of these cancer drugs could be one of the primary reasons so many of these drugs fail? I think you had said upwards of 97% fail in clinical trials. Is that a fair statement? 

13:31
David Li (Guest)
I think that is a very significant contributing factor here is that when you have fundamental misunderstandings you have a few different failure modes. One is that you leave out opportunities to expand into new indications. Two is that perhaps you're going into the wrong indication or the wrong patient population with the wrong biology being targeted with your molecule and thus this really allows you to or does not empower you to have a clinical trial that is set up for success. Before I mention the comments a little bit about why that's the case, about why mechanism of action is so difficult, maybe to add a few comments there. It really is. As we've spoken with more and more drug hunters in the industry, many who've spent the majority of their careers doing this, it's kind of an unspoken truth that finding mechanism of action and determining it in preclinical studies is kind of like reading the tea leaves. 

14:32
David Li (Guest)
You have some binding assays, you have some other biochemistry assays, you have some other kind of orthogonal signals. But it's very difficult to kind of pin down what is the true mechanism of action? And it keeps drug hunters up at night. And maybe we missed a mechanism that is actually the true reason for why this molecule has anticancer effect. Or on the flip side we miss something here that may actually contribute tox and therefore be another contributing reason for why the molecule downstream in the clinic is not successful. And so again, this kind of brings back to the main point which is we think mechanism is really important and really to be able to get it right and pull that mechanistic understanding forward to preclinical studies when we can do something still with the molecule and to adjust it. It is really important for us to have a much better chance of success when we get downstream. 

15:24
Neil Littman (Host)
And so I want to jump off then and understand more what you're doing at Meliora and understand exactly how you're bringing to light understanding some of these MOAS. But before we do, I always love hearing about the origin story of companies so I'd love to hear the sort of the origin story and maybe even just backing up. I know you started your career in healthcare, investment banking, you were at Goldman, you moved into private equity at KKR. I had a similar career path. And so I'm just personally curious what led you to leave the world of finance and move into operating roles in very young early stage startups? 

16:01
David Li (Guest)
Definitely it's definitely been a journey. So I'll rewind a bit here and start with a bit of my personal background and then segue into the launching of I in my undergrad did a dual degree that was a life sciences cell biology degree as well as a finance degree and as you mentioned started my career off investment banking at Goldman. Incredible experience and learned a ton in terms of being the first job out of school and being able to sit with CEOs of biotech organizations making really transformational decisions and important decisions for the lifespan of that company and during that time then also really started getting exposed to further to the industry of biotech. So at that point I kind of given my undergrad experiences as well as that first experience I knew I wanted to be in and around the life sciences space. When I then went to the investigative side at KKR, I realized over the long span of my career, I really wanted to look back and be able to say, that is a product that I got to a patient. 

17:06
David Li (Guest)
That made some difference. Something I can point to a team that I work with that I was really proud of in terms of that impact that I'd made. And so in early 2015 I decided to leave behind the world of finance and transition to a very early stage startup that was based in San Francisco. The company was called Benchling. Benchling is Life Sciences informatics? I was employee number six, the first business hire and came in to help lead the go to market motion sales marketing BD really an incredible experience at Benchling because that company really had over the course of the next three and a half years started a finest footing had inflection point in terms of revenue and customers and really product market fit and I personally learned a ton about early stage startups how to manage teams, how to hire how to unfortunately let people go when it wasn't the right fit, how to build culture and do all the things that are really important in the early stages of small stage company and in order to scale it. 

18:14
David Li (Guest)
So over time, for that time or around 2018, I really started thinking software is great business model, SaaS is an incredible business model, which is what Benchling was though. I really wanted to get closer again to the original goal, which is to get closer to patients. And at that moment I decided I wanted to dive further into life sciences and joined a again really early stage life sciences company that was called Everest Detection. This was a cancer early detection liquid biopsy company for specifically lung cancer and this was really the first exposure and deep dive into oncology and what was happening in that world. Over the course of the next several years I did a whole bunch of different things at Everest including leading clinical operations, setting up research labs, one in Manchester UK where were sourcing a bunch of clinical samples, one in San Francisco, and again, really just thinking through where in oncology I'd be excited. 

19:11
David Li (Guest)
It was around this time then and kind of just the starting of Meliora ran across the work of my scientific co founder, Jason Shelter. And as you said, that December 2019 paper of these mischaracterized mechanisms was a bit of a controversial one. So it did actually get some press in New York Times and other publications and the popular press. And so at the time, I was thinking through, were there other opportunities, really, for early stage investing and deploying some capital as angel and had the chance to link up with Jason? At that time, and this was already kind of late 2020, jason and our technical co founder Joan Smith had already been thinking through, how do we take the approach of making a computational method for mechanism of action deconvolution rather than some of the wet lab approaches that Jason and his lab had pursued in publishing that 2019 paper? 

20:10
David Li (Guest)
And this was born really out of necessity, COVID obviously happened in early 2020. Wet lab work was really shut down, everyone's sitting at home. And so both Jason and Joan started thinking through, well, could we develop some ML driven ways, computational ways for identifying mechanism of action? And that really then started thesis of let's leverage our understanding of mechanism to develop better cancer drugs with higher probability of success when we get in the clinic. 

20:40
Neil Littman (Host)
Yeah, it's a great founding story. And just a quick aside, Benchling is in my anti portfolio at Bioverge. It's one that I unfortunately missed out on early on. I had a chance to take a look at it and just for whatever reason didn't work out. But that must have been a really incredible experience. So David, let's then dive into what you're doing at Meliora today. And so you mentioned a little bit about the platform. You're integrating this machine learning based platform to explore the complexity of cancer, really in an effort to get a clearer view of the mechanisms and the activity of drugs that they target. So let's sort of peel the onion back. How does the platform work? What are the inputs? What are the outputs that you're looking for? Walk us through sort of the basics. 

21:29
David Li (Guest)
Sure. The platform is essentially a two step process at the highest level. The first step is we have built a very comprehensive mechanism of action atlas, or anchoromics as we're calling it. It's essentially a reference database for how different mechanisms of action are impacting cancer cell biology. And I'll give you an example in a minute here to kind of really bring that to life. But let's talk about a target. Let's say cancer targets such as EGFR. What does EGFR inhibition and all its downstream effects look like in terms of its impact to cancer cell biology, as measured through different modalities and bioassays, such as the transcriptome or the proteome or methylation signatures, or also in? Addition to phenotypic outputs like cancer cell line growth inhibition, we are still using cancer cell lines as the model system for the time being. We do intend to evolve this to other models that are closer to human over time. 

22:32
David Li (Guest)
But we do think there's quite strong signal already in cancer cell lines. And what we're essentially doing then is we're creating a very comprehensive totality of different signatures for different mechanisms of action, of course, much beyond just EGFR, right? And we're doing this in one of two different ways to create this reference database. The first is for certain targets like EGFR, there are known small molecules that are potent and selective for that target. There's a variety of different EGFR inhibitors out in the market and we can just use one of those to generate a EGFR inhibition signature in our database. Now, what about the vast majority of targets that no such small molecule exists? And for those, we create what is essentially a genomic proxy for what a small molecule inhibitor may have looked like in terms of its impact to cancer cell biology. And these genomic proxies are really what it breaks down to, this whole genome CRISPR knockout and in certain instances RNAi to be able to create again this kind of proxy for what a signature from a small molecule inhibitor may have looked like. 

23:46
David Li (Guest)
Now, the secret sauce here is several fold. One is the signatures in of themselves are of data formats. That's a proprietary type because as you can imagine, cancer cell line growth inhibition is very different from transcriptome and or methylation signature data formats. There's a lot of irregularities that need to be standardized in order to create a signature. The second technical challenge is that genomic proxy signatures and CRISPR knockout signatures are inherently a little bit different from small molecule perturbation signatures. And we need to do some normalization, some standardization and other adjustments in order to put those two different types of signatures in the same comparative space in the same comprehensive reference atlas. So all that is step one. Step one is let's create a very comprehensive mechanism of action atlas. And then step two is let's then use this mechanism of action atlas as a prediction engine for what the mechanism of action of a particular molecule of interest actually is. 

24:54
David Li (Guest)
And the way we do that is we'll take a molecule that we're trying to better characterize and we will use it to perturb a panel of cell lines, our model system, and generate a set of bioassay outputs, the same ones as previously mentioned, that comprise the signature of that molecule of interest. Then this is where the ML comes in. We will try to pattern match how similar does the molecule of interest signature look to any given reference signature in our mechanism of action reference database or Atlas. And the closer the match, then the greater the confidence that we can make a computational prediction that this molecule is actually acting via the Matching Signatures reference mechanism rather than any other target or any other mechanism that it may have been developed against. So in totality, here what it actually really is. This is essentially a modern 21st century phenotypic screen or fingerprint. 

25:55
David Li (Guest)
Instead of just looking at, say, a single modality, such as imaging or another single modality, we're looking at the totality of molecular signals under the hood of what this molecule is actually doing and then deciphering what that means in terms of the mechanism of action for the molecule. 

26:14
Neil Littman (Host)
David, how much of this are you doing in silico, and at what point do you move to the wetlab? 

26:21
David Li (Guest)
Absolutely. We are bootstrapping a lot of the initial data off of some public databases, but we are also supplementing our own data and proprietary data to be able to make a more comprehensive mechanism of action atlas. And so to directly answer the question, to create the atlas itself, we've had to do some amount of wetlab work in order to create the reference signatures. But then when you're actually trying to determine the true mechanism of action of a molecule of interest, you can do either one of two things. If that molecule has already generated annotation data, let's say it's coming from an existing data source from a proprietary company's library or some public library, then you can use that and feed it into the computational platform and be able to have it spit out what we think is a rank sorting of what we think is the mechanism of action in other cases where you need to kind of characterize a completely novel molecule that has no existing annotation data that's will then run it through again a panel of bioassays to generate its own signature and thus be able to enable this particular approach of fingerprint matching and mechanism action calling. 

27:39
Neil Littman (Host)
And then let's sort of take this one step further. How do you move from the prediction in the atlas to the actual development of a drug? 

27:50
David Li (Guest)
Yeah, I think that's a great, kind of great question, because it's really the so what if you can identify and make a prediction of the true mechanism? How does that really accelerate drug discovery? I think one of the examples in our pipeline would be a good case study of this approach. And so I'll share in a few details here. This is a molecule that was originally developed by a Japanese oncology company. And that company brought that particular molecule forward as a top kinase inhibitor. And that molecule was actually brought into a phase one study and found that it was relatively safe. It wasn't showing too many significant adverse events, but it also wasn't showing over the top signal and efficacy either. When we looked at this molecule and ran it through our computational platform and conducted our fingerprint matching approach to the molecule, we determined that it made a prediction that this molecule actually was not a top kinase inhibitor, but we predicted that it was a CDK eleven inhibitor. 

28:57
David Li (Guest)
Now CDK eleven is a relatively novel target in the CDK family. CDK family obviously has some very large drugs, CDK 46 inhibitors, multibillion dollar drugs in many different indications. But we found that this particular molecule had a strong signature match with CDK eleven at that point. Then we said, okay, let's test this molecule and see if it actually does have strong activity against CDK eleven, at least from biochemistry assays. And so we ran an IC 50 against that molecule for CDK eleven inhibition and found it was very potent in the range of sub 50 nanomol or IC 50, which is already a quite potent range with the original company not developing against its target. And so then at that point we thought maybe this is a point where we actually tested in some xenograph mouse models and put the molecule with no further changes into a xenograph mouse model, found some promising efficacy and good safety signal and at that point decided this is potentially the start of a new program. 

30:05
David Li (Guest)
We then at that point added our optimization chemistry and started making changes to the molecule so that we could one, optimize it further against CDK eleven inhibition and two, improve some of the PKP properties and other drug like properties of the molecule. And through these changes, then we arrived at novel chemistry space. And so we created a novel chemical entity, filed a provisional patent for that molecule. And we spent, up to this point, kind of less than a year and less than a half a million dollars to reach a stage where we're solidly in lead optimization, fast approaching a DC nomination. And so I think this is a very nice case study of the approach and what it can do. One, we can computationally identify the mechanism of action of a molecule of interest. That identification of a scaffold really would be able to accelerate the kickoff of a program against a novel high value cancer target. 

31:08
David Li (Guest)
And then we can progress expediently and efficiently through the preclinical drug discovery cascade. Because of our knowledge of the mechanism of action, Now I would kind of note here that everything I've talked about so far is really about there's elements of scale and speed and cost efficiency. But for us the biggest value proposition is not necessarily that scale and speed is kind of the key contributing factor here. For us, it's knowing what the mechanism of action is such that then we can develop the right biomarker strategy to identify the right patient population, that we should put this molecule in the clinic. What we don't want is we don't want to put a molecule that we don't really have a very tight hypothesis of a translational target that we'd want to be able to hit in a particular patient population where that target and that mechanism is. 

32:05
David Li (Guest)
Going to have an outsized effect on cancer and therefore we really want to be in a place where we have this very tight hypothesis around the patient population and be able to drive it in the clinic and thus in our view really meaningfully improve the probability of success. 

32:22
Neil Littman (Host)
David, I want to come back to your business model in a minute. I think that was a great case study that you highlighted. So I want to put a pin in that and just pick up on that sort of that last sentence that you obviously we don't have a crystal ball and you need to get in the clinic to actually figure out the real results. But any sense of how what you're doing might improve the success rate of clinical candidates. 

32:48
David Li (Guest)
This is the key question and certainly I will admit freely that we're not yet in the clinic and therefore we do need to get to this clinical value inflection point. What I will say, however, is that for programs that have historically, that have a biomarker, a clear patient population that you're trying to drive through and then ultimately a very tight hypothesis of the target engagement by that particular molecule from a pharmacodynamics perspective and then a clinical population where there is a need that is demonstrated with a real clinical opportunity that triumvirate and kind of linking of opportunity to molecule to target biology, clinical opportunity to molecule to target. Biology is one that historically has had much higher success rates. And that is an area where we think we really that's kind of the future where we want all cancer drug discovery to be. Although given the current tools, we just simply can't do that. 

34:00
David Li (Guest)
And that's why you see in the current many trials in cancer drug discovery now you'll still see kind of all comers solid tumors looking for any subpopulation in patients that would show the outside signal efficacy signal that enable a wide enough therapeutic index of efficacy to talks to enable this molecule to be approved. And so for us this is really where we'd like to transition and really rethink drug discovery and oncology in particular. But we think we need some ML tools and computational tools to enable us to do that. 

34:36
Neil Littman (Host)
So David, I want to go back and pull in that thread around the case study that you had mentioned. Is the plan for Meliora to build your own platform of cancer therapies or are you pursuing a business model where you're really looking to partner with other companies and let them perform clinical development and commercialization or is it some combination thereof? 

34:58
David Li (Guest)
Right, well, we do think we are a platform biotech company and what I would say is our North Star is that we want to take therapies to patients in the clinic and beyond. Now we think there's a lot of different ways to get there. Certainly partnerships is a part of that. If you truly have a mechanism deconvolution platform. There are so many different ways to monetize that value, whether that might be there may be molecules that a partner does not particularly understand what is happening yet because there might be conflicting biochemistry and genetics data. It's a bit unclear what is this molecule actually doing and therefore where should we go in terms of the clinical development that's an opportunity for us to deploy our mechanism decompolition platform. I think in other instances what we can do is start with a library of chemical scaffolds and molecules and be able to pick out what the true mechanisms of action are for all of those molecules. 

35:59
David Li (Guest)
And again, because we're doing it computationally, we can do this at scale and then find the scaffold target pairs that we are really excited about that gives us the best chance of success downstream in the clinic that we want to kick off new program development and could do that in partnership. But I think, again, circling back to where we ultimately like to go is we do believe that the best way to have the most impact for the patient is to develop the and take best advantage of the platform. Value proposition around mechanism decollution is to develop the assets into the clinic ourselves and also bring them to the patient. 

36:35
Neil Littman (Host)
David, let's switch gears here a little bit. I want to talk about the finance side of things. And so you were able to raise $11 million in seed financing in September of 2022. This was a very difficult environment for startups to raise money, particularly biotech startups. What were those discussions like? What was it like raising capital in this current environment? And what, if any, advice do you have for other entrepreneurs who are out there trying to raise capital at the moment? 

37:10
David Li (Guest)
Certainly we're definitely in some very interesting times in biotech. I've heard from many in the industry it hasn't been to this extent in kind of 1520 plus years. And so we're really at a unique time. I'm not sure we did anything completely unprecedented or anything like that in terms of our raise, but I will say that we had some lessons. The first was this is kind of the first point you must get correct is just the story. What is the problem you're trying to solve and how does your solution directly solve that problem that the industry has? I think oftentimes, especially when technology is really the actual science is really interesting and the technology platform is really compelling, there is a bit this kind of conviction around the end market and the end problem being solved that is left open to interpretation. And what our job is as entrepreneurs is to draw that line from technology to market problem and make it very clear how exactly we get there. 

38:15
David Li (Guest)
So I think that's absolutely first is that from a strategic perspective, you really got to make sure your story is really compelling. First, tactically, I would say that in early stage fundraising, and this is a lesson I think I've picked up not just from this particular process at Meliora, but also at Venturing and previous entrepreneurial pursuits. It's this early stage, what you're really selling in terms of equity in the companies you're selling trust. The investor has to trust that you will be able to create a massive company out of this platform, out of this technology, out of this very early start. And there are a lot of different ways to gain trust and to establish trust, whether that might being an expert around a particular market or a particular modality or a particular portion of the industry or a particular technology sometimes. But also it can be gained through consistent execution and consistent kind of interaction. 

39:20
David Li (Guest)
Really it takes time to build trust. And if I had to go back and kind of give myself advice, if I were to do this over again, I would say there is sometimes kind of a hesitancy to engage investors of broader community until you think you're ready. And I think that sometimes will pull you back a bit when really what you should be doing is really honing in on your story, but engaging with the wider investment community early and often to really hone in. On that story and figure out how you can build trust with them so that when the times comes time to raise, you're able to do so. But it certainly is a tricky market out there and I really do feel with all the entrepreneurs that are trying to do something great in this environment all good advice. 

40:07
Neil Littman (Host)
All good advice. Well, David, we could probably talk for another couple of days about some of these topics. I do want to be cognizant of your time and wrap up with one final question, which is one I often like to ask, but I think it's really pertinent, particularly in this environment. You mentioned you're building a platform company. Obviously you've integrated a stack of technology, machine learning and other computational aspects that you've described during the show. What is your view on tech bio versus traditional biotech? Do you think there are some key differences or is there one that you lean into more than the other? 

40:44
David Li (Guest)
It's an interesting bifurcation. I don't know if it's necessarily kind of one axis that really separates the two. One thing I guess I will bring up is around just a tipping point around culture. And the culture I'm describing is just kind of where is the value generation of the entire company? Is it really around the data and the computational platform and is that driving kind of the value inflection and being the engine for the company? Or is the data platform and kind of computation and the data team really seen as a support function? And what I'll say is I think two things can simultaneously be true. Data and more advanced computational strategies. Will eventually permeate every single step of drug discovery from target identification to validation to hit ID and so on and so forth. So I think eventually there's no doubt about that's just the onward progress of using better technology over time to support the end goal. 

41:53
David Li (Guest)
But I think there will always be this kind of spectrum of folks where there will be kind of the data and computational platform center being almost kind of a support or cost center of sorts versus ones where they are really the front and center, they are the engine driving the value for the company. And I do think there will place for both over time, but in our view, for us, this would not be possible without just the machine learning capabilities to create the signature to make pattern recognition matches. This is really enabled by kind of a sign of the times of having the data available and having the compute capabilities to pull out signal. So we definitely think that there's a lot of value there, but I'm sure over time there is space for both. 

42:40
Neil Littman (Host)
I agree. And I think there's an increasing convergence too between sort of where people draw the line as well and I think culture is actually a big part of that. So David, with that, I think we better wrap up and I'd like to say a big thank you for your time today and joining me on the show. 

42:56
David Li (Guest)
Awesome. Thanks so much Neil, it was really enjoyable. 

43:01
Danny Levine (Producer) 
Well Neil, what did you think? 

43:03
Neil Littman (Host)
I thought that was a really great conversation with David. It was a pretty wide ranging discussion around our current lack of understanding of the MOA that surrounds a lot of the drugs that we are bringing forward and that are even approved today. And so I think his thesis is spot on. If we can have a better understanding of the MOA, we'll be able to more precisely target specific cancers and that should increase the probability of successfully developing drugs. Right. And so you heard David mention this 97% failure rate for novel cancer therapies in the clinic. Even if we could move that a couple of percentage points, that's huge, right? I mean, we don't need to move that 97 down to 50. I mean that would be great. But even if we move it down to 95 or 93%, that's huge progress. And so I think what Meliora is doing is giving us the ability to really make a dent and increase that probability of success. 

44:00
Danny Levine (Producer) 
How about the approach the company is taking to tease out the mechanism of action and the use of machine learning? 

44:08
Neil Littman (Host)
Yeah, I think it's really a powerful combination. You heard me ask David the question about their platform and how do they move from the prediction of their anchor Omics Atlas to developing an actual drug. And so you heard David talk about the mix between some of the in silico work that they're doing, the wet lab work that they're doing to validate the platform. I thought the case study that he provided about that drug that they're working on from The, I think it was a Japanese pharma company initially. Right. I think that's super interesting. And so I think that all of these things are going to be combined into hopefully moving the probability of them developing a successful drug candidate by at least a few percentage points, if not more than that. Obviously we're not going to know until they're in the clinic with a candidate and they're not quite there yet. 

44:59
Neil Littman (Host)
But it feels like they're doing all the right things to increase the probability and give themselves an edge in developing novel cancer therapies, which is to me really exciting. Right. At Bioverge, we're always looking for companies that have some sort of edge in better predicting how patients are going to respond to therapies to increase the probability of success. And Milior is not in our portfolio, but that's exactly what they're doing. And so I really believe in sort of the platform that they've built. 

45:26
Danny Levine (Producer) 
David mentioned that there are many ways it could choose to capitalize on the value of its platform. What do you think of the different opportunities and how does it maximize the value of the atlas it's building? 

45:40
Neil Littman (Host)
Yeah, I mean, that's a question for all platform biotech companies at their stage, right? I mean, you could have a model where you're helping others develop drugs and that could be a very nice business model. You're bringing in near term revenue, non dilutive sources of right. You can potentially get downstream milestones and royalties on drugs, so that can be a very nice business model. But the other way is obviously to use the platform to develop your own drugs. Right. So I really appreciated what David said about their North Star. Right. And so I think that really helps inform what they're doing. And so getting more drugs to cancer patients and increasing a probability of successfully developing those drugs is their North Star. How they go about doing that is important, but whether they have the partnering business model or they're developing their own therapeutics or some combination thereof is sort of a means to an end. 

46:45
Neil Littman (Host)
And so companies at their stage, I think, can and probably should pursue both. Right. And it sounds like that is what they're doing to some extent. 

46:55
Danny Levine (Producer) 
The high cost of drug development has long been about high cost of failure. Much of what Meliora is seeking to do is cut the number of clinical failures. We've seen so many technologies designed to accelerate drug development and improve the safety and efficacy of therapies. Despite this, we tend to get some interesting drugs, but not really move the needle on drug development. Do you think that's going to change with AI approaches like this? 

47:28
Neil Littman (Host)
Well, Danny, that's the billion dollar question. The hope is that it absolutely will. I think we're already seeing some of the first waves of AI developed drugs enter the clinic. Recursion recently received a $50 million equity investment from Nvidia, the big GPU manufacturer. So I think there's a lot of data points that are helping to validate AI being applied to the drug development and drug discovery process. I think over time it will absolutely pull out costs and hopefully increase the efficiency in which new drugs are developed and move through the clinic. This will take time though, right? This is not going to happen overnight. It's been happening for the past decade. We're now starting to see some of those efforts bear fruit with clinical candidates. I think it will accelerate over the relatively near term, but it's not until we're going to start seeing how these drugs perform in the clinic to really see how the landscape is changing. 

48:37
Neil Littman (Host)
So I think we have to wait and see, but I'm very hopeful. I think things will change. I think AI will play an increasingly larger and larger role, not just in the early discovery process, but sort of throughout the lifecycle of developing novel drugs. Well, until next time, thank you, Danny. 

48:56
Danny Levine (Producer) 
Thanks for listening. 

48:58
Danny Levine (Producer)
The Bioverge podcast is a product of Bioverge, Inc. An investment platform that funds visionary entrepreneurs with the aim of transforming healthcare. Bioverge provides access and enables everyone to invest in highly vetted healthcare startups on the cutting edge of innovation, from family offices and registered investment advisors to accredited and non accredited individuals. To learn more, go to bioverge.com. This podcast is produced for Bioverge by the Levine Media Group. Music for this podcast is provided courtesy of the Jonah Levine Collective. All opinions expressed in this podcast by participants are solely their opinions do not reflect the opinion of Bioverge, Inc. Or its affiliates. The participants'opinions are based upon information they consider reliable, but neither Bioverge or its affiliates warrants its completeness or accuracy and it should not be relied on itself. Nothing contained in accompanying this podcast shall be construed as an offer to sell, a solicitation of an offer to buy, or a recommendation to purchase any security by Bioverge, its portfolio companies, or any third party past performance is not indicative of future results.