Big Pharma Fails 50% of the Time in Phase Three. AI Can Fix That | Vin Singh, BullFrog AI

0:00The point of this is to increase likelihood that a clinical trial will succeed. Considering it takes 10 to 15 years and 1 to 2 billion dollars to develop a drug, it's quite shocking that Big Pharma fails 50 % of the time in Phase 3. Are you concerned about competition from Big Pharma who may be developing? We discovered the biological drivers of those diseases. We believe we've discovered the genes that drive depression, bipolar, and schizophrenia. From a one-of-a-kind data set that does not exist anywhere in the world. It seems that you would be an acquisition target by a big pharma. Let's see if we can get one of these drug discovery AI deals done, or target discovery deals done.

0:45Introduce yourself to listeners, how you got to Bullfrog. Craig, thanks for having me today. Great to be here. um so uh you know my name is vin singh i'm the chairman founder and ceo of bullfrog ai um we are a publicly traded company under the symbol bfrg on nasdaq so obviously there are certain things i won't be able to say or questions i might not be able to answer but um looking forward to it so a little bit about me i have almost 30 years of industry experience uh life sciences biotech um i'm a biomedical engineer by training i'm a three-time founder of investor-backed companies two of them have gone public uh the first company i co-founded is a company called max site uh similar as mxct as a cell therapy systems company when they went public on nasdaq several years ago there i think the market cap was around 1.5 billion.

1:48Obviously, biotech's been beaten down pretty hard over that period of time. After that, I founded a company called Next Healthcare, which was at the time the world's first adult skin cell and stem cell banking company for regenerative medicine applications. And that company's still operating and doing some different things today. And then And about eight years ago, I founded Bullfrog. And the reason I founded the company was I was, you know, very surprised by the high failure rates in late stage clinical development. And, you know, considering it takes 10 to 15 years and one to two billion dollars to develop a drug, it's quite shocking that big pharma fails 50 % of the time in phase three.

2:36It doesn't, it didn't make sense to me. and I also realized if I could make a difference, there's a lot of people that will benefit from drugs that maybe otherwise wouldn't. And I think the trickle-down effect would be, you know, drugs would be more affordable in general if you have more success, right? Because somebody's paying for all those failures, right? And it's unfortunately, it's the healthcare system that's paying for it. So, you know, I started the company and that the original focus was, why don't we try to rescue failed drugs? Cause there's like a ocean of them out there. Obviously, you know, it's, and I still something that we're interested in, but when I had started first, I realized, wow, that's going to require a tremendous amount of capital.

3:26And the funding environment is not what it is today, right? Where many billions are just flying around, right? Back then, you know, eight years ago, AI, artificial intelligence, was not really, you know, a well-known term. You know, now it's like every human being on Earth knows what it is. So it's just we're in a different world. But so I started the, you know, the company, you know, I realized like, you know, technology is the key here. Right. Yeah. And I had relationships with Johns Hopkins University. I did my MBA there and I was a mentor in residence as well. And, you know, I was hunting around and it turned out that they had a subsidiary called the Applied Physics Lab or APL.

4:17and um you know uh one thing led to another and you know discovered that hey this is a multi-billion dollar subsidiary that has incredible technology and does very sensitive defense work um and like you can literally can't even get in any of the buildings there i mean it's a pretty impressive place so they they develop all kinds of amazing technologies things that the average person's probably read about or seen on TV related to like space exploration and robotics and things like that. And, and, uh, they had spent years developing an AI platform, you know, before it was like the thing to do and, uh, had deployed it in different sectors.

5:03So we were able to secure, uh, worldwide exclusive license for drug development applications. So right out of the Gates, we had blue chip technology. This, um, an APL, they, you know, they have like 5 ,000 engineers and scientists there. This is a big place. And this tech won innovation of the year, uh, at APL. So, you know, any, so we, we had this tremendous technology. Um, you know, it was originally designed for different applications. And as we know, biomedical data is very different than, and autonomous vehicle data or energy data. So we went through some R &D for a while and then ended up with the platform that we needed, did some more work with them, and then added our own secret sauce.

5:55And here we are today with a very unique and powerful ensemble of technology and capabilities. and uh you know craig everybody says oh you know we have the best technology and of course i'll say that too but but now we're getting you know feedback we're hearing from you know big pharma for example that knows what's out there they have their own teams they've done deals and they're you know they're telling us that nobody can do what you guys can do which is fantastic uh so obviously we want to, you know, next step is, well, let's do, let's work together then, if that's the case. So, so that, you know, that, so that's where we are now.

6:39We're, we're, we kind of followed with our market forces and where those forces pushed us was toward drug target discovery, right? That's the very first step in drug discovery. What is the right target? Okay. And then the last step, late-stage clinical trials, phase three. So that's where we have really, you know, had our experiences and made a difference. But I just want to be clear, though. Our technology is completely agnostic. It doesn't, we don't care what stage of development or what disease. You know, the key is the data. and you know we know that we can make discoveries and uncover insights and make predictions that are you know very few if anybody can do so so that's that's the story that's how we got where we are today okay and when you talk about uh the technology of a bullfrog's technology are you talking about um on both ends on drug discovery and in the phase three trials uh are is it a platform or are they tools that you're making available to uh either big pharma or small pharma drugs and yeah just talk about the two ends is it a platform are they tools what what is the product So right now, it's a platform, but the customer, client, they have to send their data to us.

8:23Let me zoom out a little bit for you. I want to talk about the whole platform, all the solutions. So we are now in a situation where we can address end to end, right, everything that needs to be done, right, using AI for drug discovery and development. And, you know, we've launched, we launched a new solution yesterday, actually. We put out, you know, press release on that. And then we had launched one several months ago. But I'll just walk you through them quickly and then you can ask some questions. But so the first thing we realized is, you know, the companies that we were working with and we just know in general, there's a lot of messy data out there.

9:09There's a lot of data, but it's a mess. And with the like rapid adoption of AI, right, these and these, you know, with chat GPT and Claude and all these companies, you know, we were like, wow, you know, that's great. but nobody really has data that's ready to be ingested by an AI, right? And so you're not going to extract the value that you should be out of the AI. And so through one of our experiences working with this phase three stage pancreatic cancer company, so we went through this exercise with them. You know, they presented their data to us and we were just like, wow, what is this? you had tens of thousands of pages of pdfs with handwritten notes i mean like it was a mess so we developed an algorithm and pipe algorithms and pipelines to accurately convert all of that data all the handwritten notes and everything into clean structured data and going through that exercise we realized that if we can do this with a gigantic mess of data like this we can literally do this for anybody else probably in any industry to be honest because other industries data is far simpler than biomedical data um and when we said all right this is we're going to call this bf prep for data prep so we launched this several months ago and as we've done more homework, we realize even now more than 60 % of companies out there don't have AI-ready data.

10:48But they're integrating these AI solutions into their IT systems. But let's just see what happens. But they got to prepare their data. So we can do that. Okay. That's a very powerful AI tool. Then the next step is the analytics. And that's something we call BFLEAP. And that's really what I've been telling you, you know, for the past 10, 15 minutes here. That's where we make the discoveries. But sometimes you'll make a number of discoveries, like let's say you discover a number of drug targets or a number of biomarkers, right, or some other prediction. That's where we just launched BF Arenas, okay?

11:32And that says that the purpose of that is look at all these options and help me figure out what is the best one for us to move forward with. And, you know, that's a really critical... Just in terms of molecule, is that what... It could be the molecule. It doesn't really matter what it is. Think about any time you've done some quantitative analysis. Let's say you've done a risk analysis and you have all these risks. You arbitrarily assign weights. It's like, okay, this risk is worth 20%. That's worth 10. This is worth five. There's a lot of subjectivity in this, right? And it's like, okay, how severe is the risk?

12:15What's the likelihood? You know, you may have some data. But with arenas, you take that subjectivity out of it, right, by using large language models and other tools and incorporating massive amounts of information so that you can, you know, truly make the right decision And, you know, a lot of these AI tools are black boxes that spit out answers, right? How do you trust it? How do you know, oh, should I do this or not, right? So we are adding that layer on top, which we think is going to be, you know, a fantastic piece of the entire continuum, right, of solutions that we're providing. um so you know with all of this what this speaks to is we are in the ai innovator category craig we're not like these biotechs that have added ai and you know people companies out there that are using the open source tools and putting their own wrapper around it we are innovating right we're pushing the industry forward we the things that we are doing you know we're pretty sure in almost all cases nobody else can do them uh but we're still this tiny company we're under the radar right so that's sort of the the tug of war that's going on but we're driving toward getting deals done now with pharma our goal you know we want to be a revenue generating company we're not a biotech we're not developing drugs for the long term we leave that to our customers right and let Let me just ask about the middle part of big BF LEAP.

13:55I thought that was specifically for precision oncology. That was one of the applications. So I can tell you a little bit about what is BF LEAP. So the core of it, the heart of it, is something called scalable graph analytics. Okay. That's what comes from the Johns Hopkins Applied Physics Lab. we've now over time added causal AI capabilities. And let me explain what that is. I think your audience will be really interested. So typically with AI, you're looking for patterns and relationships that exist. That's how deep predictions are made. So we've added something now called causal AI. We've developed this ourselves where we can generate 6 million models in 30 minutes.

14:45And with causal AI, you can not only identify the relationships and patterns, but you can determine the magnitude and direction of the relationship. So why is that important for our industry? Biological systems are very complex. So if you want to try to determine, say, hey, what gene is driving a disease, driving these disease pathways and ultimately leading to the disease, right? Having that capability is an advantage, and we've used that, and that's helped us do some of the things we've done. For example, our partnership with the Lieber Institute for Brain Development, we had access to all of their data.

15:32They have many thousands of post-mortem brains for neuropsychiatric patients. and the world of neuropsychiatric you know not well not only diagnosing but developing drugs very challenging space a lot of subjectivity there it's about behavior and symptoms right well through our analysis of this incredible data we had exclusive access to it we discovered the biological drivers of those diseases. So now you add that to behavior symptoms, you have a precision neuropsych approach to these. So we believe we've discovered the genes that drive depression, bipolar, and schizophrenia. And that is, so those discoveries, now we've gone to market.

16:24Now we're talking to big pharma. Hey, let's work together here. We've discovered the drug targets and like i said in the beginning i didn't really say it but getting you know the wrong drug target is the reason a very significant number of drugs fail okay because with these causal networks we can show it's like it's like the root cause of the disease we we can actually illustrate it hey this gene has no parents everything is downstream from it this is the root cause right Right. It's not the gene way at the end that then, you know, you know, these genes turning on and off causes that it's this root cause.

17:11A lot of times when drug targets are discovered, it's downstream somewhere. It's not the root cause, which is why the drug fails, you know, somewhere in the middle of development. So being able to pick that right target from the beginning, it's going to save the industry lots of time and lots of money, right? Because not only could you use it to discover new targets, but hey, if you're a company that's already, you think you've discovered the right target, well, we can tell you if you did, right? And I, you know, so I think this is something the whole industry is going to eventually really appreciate and investors too.

17:50I mean, you know how high-risk biotech investing is, right? Yeah, sure. Just a question about that. But diseases don't have necessarily a root genetic starting point. Not necessarily. Yeah, a matrix of combinations of different things. So how can, yeah, maybe take one example. I'll tell you, in taking one example, what caught my attention when I heard about you guys is BFLEAP has identified patient subgroups with a nearly threefold increase in mean overall survival of pancreatic cancer. And pancreatic cancer is like one of those boogie men that people might be terrified of. Because, yeah. So could you use that as an example or is that a little different than?

19:03Well, it's like I think we're talking about maybe two different things here. But let me first say we are our platform can ingest any type of data. Right. You know, we were just talking about gene expression or genomics data. but it's true multimodal data imaging data demographic data environmental data lab data all the omics data you can imagine so you're right Craig in a lot of cases it's much more complicated than just a gene right but that's why we're able to bring all that information in and that's the precision medicine approach because then you can really identify the true genetic and non-genetic profile of a patient that will maybe in this case best respond to a pancreatic cancer drug.

19:54So yeah, in that example, that's when we had this mountain of data, and we were able to identify a patient subgroup and biomarkers that increased the overall survival from like two months to six months, which is, you know, for pancreatic cancer, that's a big deal. but it also speaks to the fact that hey the answer is in there like there is a you know that's a precision medicine approach if you're a patient with this particular profile this drug will benefit you right so the whole industry especially with ai it's almost being forced in that direction i mean we've been hearing about precision medicine for 20 plus years but there's all there's always that tension because pharma wants wide nets for the patient population right they want millions of patients yeah well if you want drugs that actually work you're not going to be able to do that right you got to take a precision approach so i think with with mass adoption of ai it's going to turbo boost the precision medicine uh initiative that's out there yeah and also uh you're talking about big pharma but what's happening are there are smaller uh companies labs that are developing uh molecules or or drugs uh and they bring them all the way up uh through phase one phase two trial and then sell them to big pharma is that affecting that strata of the market those companies it's not big pharma but they're focused on on uh precision drug discovery yeah it is i i think what i'm seeing is more and more biotechs now are trying to incorporate ai from the beginning right i mean they realize it's only going to improve their chances for success, right?

21:56The question is by how much, but still in this incredibly high risk business of biotech, where your chances from start to finish are like low single digit percentages, um, any improvement is, you know, it's, it's a big deal. So, uh, yeah, we are seeing that, that that's becoming. And then if, like, if you look at other countries, China, for example, I mean, they've bolted AI into their entire biotech industry. I mean, that's it's like having electricity at this point. Right. I think U.S. has been slower probably on the biotech end to adopt that. Pharma has understood the value for the past 10 years.

22:39That's why they've been doing all these deals. And now they're starting to make acquisitions in AI because they want to own their own models. they don't want to partner on them they want to own them so they're making acquisitions or they're doing what eli lily did and said hey we're going to partner with big tech and build super computing capabilities and you know just develop our own models and so forth so anyway so that's sort of the you know some of the changes that are occurring but it's it's all in the right direction for us you know we're you know it's all good yeah and so you guys you're saying you're your disease or target agnostic yeah we are disease uh and stage agnostic yeah so uh how does a company use and let's talk about the three platforms uh how do they what's the the the the progression i mean sort of give me a use case maybe that neuropsychiatric use very interesting happy to do it i mean so the first thing is we talk to a company we basically audit their data okay what kind of condition is your data in um and we have bf prep we can prepare that data so this is a great it's a great door opener for us even if they don't want to work with us beyond that we know they got problems with their data, right?

24:09So that's the first step. Then, you know, then we can basically prepare, you know, clean and structure their data. So we can do what would take a team months. We can do it sometimes in hours, right? But certainly in days, we can prepare that data. Then the next step is, all right, what kinds of, you know, what questions do you want answers to in this data what can we help you find and that's where bf leap and the analytics comes into play right so out of that we're able to build these networks these networks show the relationships that exist in the data or you can you know identify disease pathways and you know things like that then let's say a number of predictions are made right and like an example i gave earlier was drug targets right right and let's say there's a series of genes that are we think they're the drivers of a disease okay which one should we go after out of those 10 driver genes and that's where we layer on bf arenas so bf arenas effectively takes each of those you know selections and creates a competition between them which is that that's really the secret sauce that's the approach yeah um and brings in lots of other information and then what you're presented with is a recommendation that you can have a lot more confidence in okay vf arena is the thing go go for target one and then six and then nine you know so it's like a priority ranking so anyway and i think that's going to be critical in this world of ai right because you're gonna you're the AI is, you know, you want to, first of all, you have to trust what the AI is telling you.

26:01And that's where explainability is important. And our platform's always been explainable from the very beginning. It's not a black box. Right. But then even with the explainability, okay, help me make a decision. Anyway, I think we're uniquely positioned. you know we're a small company but like super high on the innovative uh end of things and uh so now we think we got like three really valuable pieces for the industry that that are going to make a difference yeah uh the the uh at the at the last stage there uh when you said then you have these these different targets uh competing with each other are you using evolutionary AI where you're coming different models produce an output and then you take the best and combine those models.

27:02Yeah, so we're looking at different models and like I said, with our causal AI capabilities, for example, we can generate millions of models in a matter of minutes, but that's the beauty of it like you're bringing lots of information together, lots of models, lots of outputs. And then you're trying to find sort of where that common thread is to give you confidence in what selection to make. I mean, you said an interesting thing earlier that you were looking at, you were considering looking at failed drugs because clinical trials fail oftentimes for reasons that have nothing to do with the underlying drug.

27:49But the point of all this is to increase the likelihood that a clinical trial will succeed, right? I mean, ultimately that's the goal. That is the goal. That's been the goal from the beginning, to de-risk drug development, improve the odds of success, shrink the timelines and investment. That's always been the goal. And like I said, we've had opportunities to get involved at the beginning, at the end. Hopefully in the future we'll get more involved in the middle stages. And I think that's the goal of AI in general, right? I mean, it doesn't matter what industry you're in. It's to improve efficiencies, right?

28:37Yeah. So these three platforms, BF Prep, that makes the data, cleans the data, prepares the data, and then the causal analysis, that's BF Leap. and then the sort of decisions on what to focus on, that's BF arenas. Can you talk about Bullfrog data networks? From what I understand, that's sort of the, that ties them all together. Yeah, really. It's, well, it's really the outputs of BF Leap. So BF Leap is the analytics engine, but what comes out are the networks, right? And that's Bullfrog data networks. And that's where you can actually visualize these relationships, extract information from whatever's being illustrated there.

29:34So that's what the networks piece is. Yeah. And again, on the underlying AI, can you talk about the different models? I mean, you said it's an ensemble of models, but can you talk about the different architectures underlying of the three platforms? Yeah, I mean, I can't talk too much about the architecture. That's kind of a, but I can just tell you scalable graph analytics is like, is the core. And that's, you know, that's an approach that's been around for a long time. um but what we have it has is proprietary it's protected by patents some of it uh but that's really the the heart of the technology uh these other pieces that we've attached you know that that's our innovation but um yeah i want to be careful about giving away too much of our secret sauce here, but, uh, you know, at a high level, that's what, uh, what we use to do what we do.

30:45Yeah. And do you have any, uh, case studies you can talk about? Um, yeah, we do. I mean, we, well, obviously, uh, one is this phase three pancreatic cancer, um, work we did and we're actually, we'll put out a case study on that pretty soon so that everybody can read this whole story from start to finish. The other one is, you know, the work we did with Lieber, where we discovered those driver genes, these drug targets. I mean, Lieber has been around for like 15 years, and they spent many years trying to figure those kinds of things out, and we did it in a matter of months. Yeah, we'll talk about that sort of start to finish.

31:28Yeah. Yeah, so, you know, we, so, so, Libra has, I think, over 5 ,000 brains, okay, actual human brains, and the vast majority of them are for neuropsychiatric disease. They also have neurodegenerative, you know, and other related diseases, and, you know, we had talked to them about our technology. They are associated with Johns Hopkins Medicine. So they were kind of familiar with APL, the Applied Physics Lab. And they were like, okay, if that's your technology, like we want to work with you because we know how good those guys are. And so we had some other connections with them. Our senior director for CNS had worked there for many years in the past.

32:23And so, you know, there was a lot of trust there because for Lieber, look, they can work with anybody. They had to feel like they could trust whoever was going to be their partner and who was going to have access to this precious data. I mean, they spent a lot of money, probably over$100 million in many years collecting these brains and extracting tissue samples and then generating the data. I mean, a lot went into this. So anyway, so that's how it all started. And then we worked collaboratively with them. It wasn't like, all right, hey, bullfrog, take the data and let us know what you find. We met with them on a weekly basis.

33:10They have tremendous expertise there. Their CEO, by the way, Danny Weinberger, he's a rock star. He's like Nobel Prize winning type guys. He's a phenomenal scientist, physician, and leader. He was actively participating in these meetings with us. And as we, you know, each week went by and we're showing them what we're doing and you saw the eyes wide open, like, you know, wow, that's pretty impressive. And then anyway, so we got through that in a matter of months. And, you know, then the goal was, OK, these are important discoveries. OK, we know pharma is very interested in neuropsych. Right. Also neurodegenerative, but that's for another another day.

33:59uh they're interested in neuropsych we know there's been a lot of failures in in that space and uh but so let's you know let's start approaching pharma let's see if we can get one of these drug discovery ai deals done or target discovery deals done and as you can imagine it's a long sales cycle working with pharma i mean easily over a year So, you know, we've been doing that, though. We've been doing the outreach. And then along the way, trying to prove ourselves in other ways, right? Because, you know, when you're this small company that no one's heard of, you know, even pharma will look at you like, you know, who are you guys?

34:44So, you know, we did some things along the way, like the phase three pancreatic. We had some partnerships. And those were forms of validation for us, right? They said people were saying, hey, okay, if you work with those guys, then, you know, you must have something. So anyway, so fast forward to today. We've been in a number of these conversations, and, you know, we're zeroing in now on some real opportunities. And, you know, the way it works with pharma, you know, for those first opportunities, it's typically like a pilot. Like they have to kick the tires on Bullfrog, right? And if that goes well, then you get into the mega million dollar deals that we keep reading about.

35:28I think a lot of people don't realize that. They just see these big headlines and they, oh, wow, how'd they go from start to that so quickly? Well, you don't know what happened in between, right? In between is the pain and suffering and having to do these small projects, right, to prove yourselves. but uh so that you know that's that's how we got where we are today we continue to have the relationship with libra um and we look forward to you know continuing that into the future right right but but you identified a gene the root cause of one yeah in the analysis we we discovered the driver genes we call them right so we you know these are the genes that appear to be drivers of the disease right but which which disease specifically are you talking we found it for neuro for major depressive disorder bipolar and schizophrenia okay you're you take that to big pharma and you're looking for someone to take this up and and develop uh a drug to address that exactly exactly what we're saying is hey guys we've discovered very valuable targets here novel targets from a one-of-a-kind data set that does not exist anywhere in the world okay it's not public data nobody has access you know had access to that data the way we did so yeah so that's the pitch we're not going to identify the drug do drug screening or design a drug that's pharma's job we've done our job that said our plan is you know once we get a deal or a couple of these deals done, obviously there's three disorders, so we can do at least three deals, presumably, probably more.

37:29Once we get cash flowing from those, then maybe we'll advance these targets one or two stages ourselves. We don't want to become a drug development company, but we can take that target, we can then validate it in the lab, and then maybe we do a drug screening then we do the deal with pharma right right because that's not a super long time line there it's a few years right and it's not a massive investment but it de-risks the asset and makes it much more valuable so we can do bigger deals so we're trying to do things one step at a time it's like we've made these discoveries so well i was going to say and And as I said before, there's this strata below Big Pharma that is developing molecules and taking them through phase one or two clinical trials and then selling them to Big Pharma because phase three is much more complicated and expensive.

38:30Are there any of that level of company that you're talking to or interested in? That's not our focus. I mean, we're happy to talk to them. Do they have resources, right, to do a deal with us? And that's where I think their investors have to get a lot smarter and say, hey, we got to add in an AI budget, an AI partnership budget to these investments that we're making, right, so that they can take advantage of some of these opportunities. But, you know, I think I'd say like a year or two ago, we were talking a lot of biotechs. But what we realized is, you know, they didn't really have the money. They didn't really understand the value of AI.

39:17They didn't know what they were really looking for. You know, that's the thing. They kind of had tunnel vision where big pharma is the opposite across the board. Right. So we said, let's focus. You know, we're a small company. We have limited resource. Let's focus on big pharma because that's the customer we want. um so we're not opposed to dealing with biotechs uh but that's you know we just kind of made that strategic decision uh a while back yeah so where are you uh now i mean you you you sort of brought us up to date we know about the three principal platforms uh and and you've had you're you're working on this neuropsychiatric target or targets, where do you as a company go from here?