Unlocking the Complexity of Disease with Transomics and AI

Daniel Levine (00:00.398)

The Life Sciences DNA podcast is sponsored by Agilisium Labs, a collaborative space where Agilisium works with its clients ranging from early stage biotechs to pharmaceutical giants to co -develop and incubate POCs, products and solutions that improve patient outcomes and accelerate the development of therapies to the market. To learn how Agilisium Labs can use the power of its generative AI for life sciences analytics,

To help you turn your visionary ideas into realities, visit them at labs.agilisium.com.

You're tuned to Life Sciences DNA with Dr. Amar Drawid.

Daniel Levine (01:08.782)

Amar, we've got Samantha Dale Strasser on the show today. For audience members not familiar with Samantha, who is she? Samantha is a co -founder and chief scientific officer of Pepper Bio. She leads the development of Pepper's technology. She was an NSF graduate research fellow, Churchill scholar, and Goldwater scholar. She pioneered omics analysis approaches in electrical engineering and computer science, as well as biological engineering at MIT.

What is Pepper Bio? Pepper Bio builds itself as the first transomics drug discovery company. It's trying to address the high failure rate and high cost of drug discovery and development by using AI to inform better decisions early in the drug discovery process. And what are you hoping to hear from Samantha today? Pepper Bio is being ambitious, going after difficult to treat diseases like highly aggressive,

and intractable cancers. I would like to understand how it is able to use transomics and leverage AI to do that. Well, if you're all set, why don't we welcome Samantha to the show? Samantha, thanks for joining us. We're going to talk today about difficult to treat cancers and how Pepper Bio is using a range of omics data to inform decisions early in the drug discovery. So let's start with the notion of transomics.

For listeners who are not familiar with the term, can you define what is Transomics? Certainly. I'm really excited to be here today and share what we're building at Pepper with Transomics. Big picture, transomics and what Pepper is for is here to treat untreatable diseases. And so Transomics itself is a analysis approach that we often liken to the changes we've seen over time in real -time traffic data. So if you think about, you used to have a big clunky map when navigating from point A to point B.

And it gave you where roads were, but no context of what was happening on those roads. Real -time traffic data you now have on your phone, this ability to see where vehicles are and what's actually happening around you to make a more informed decision. So transomics and Pepper is doing that for drug discovery. So we're bringing a context of functional information and context of understanding of what's actually happening in a disease or when a drug is applied to treat a disease.

Daniel Levine (03:33.774)

to make the most informed decisions of how to help a patient and at the end of the day, end untreatable diseases.

And can you tell us more about what are the different elements of transomics? Absolutely. So underlying transomics, we actually were integrating multiple different omic data types. Today, that includes genomic data. So sequencing your genome, like if you sent out 23andMe a sample to understand what's there. Another data type called transcriptomics, which is basically taking the genome and it's an instruction set of how to make a next type of data called proteomics. So proteins, which...

Proteins are really critical in signaling as they're essentially within a cell, the molecular actors that get things done. So they're the molecules that either tell a cell how to grow, to move, to live, and to die as a cell. Now those proteins have different functions. They can be turned on and off. And that's then moderated by another data type we work with called phosphoproteomics. And so phosphoproteomics itself is a data type essentially looking at little molecular switches that are added to proteins.

that change their function to tell them what to do. And so by looking together at all four of these different data layers, we have the most comprehensive understanding at Pepper of what happens in a disease under study, and also what happens when a drug is used to treat a disease. And this really rounds out that picture in drug discovery of how to best help treat patients. So the genes.

Transcript to transcripts the RNA and then from RNA you go to proteins and multiple proteins there, right? So you get the genomics, transcriptomics and the proteomics. And then of course you're talking about the switches, the phosphorylation of the proteins. So how do you like in, when you are putting these together, how do you combine these? So a really key facet to interpreting all these data types when looking at a disease.

Daniel Levine (05:34.382)

is having what we call a reference data set or a dictionary to understand how they relate. So just like you use a dictionary to know what words do in language, we have a dictionary to know what different proteins, molecular actors, and genes do and how they relate to each other, since it's really critical to know relationships between them to understand how they function, how it changes in disease, what implications that has in this, what we call a global picture of molecular signaling.

to understand comprehensively what, just to give a sense of scale, thousands, tens of thousands of molecular actors are doing. It's a big picture, a phenomenal point we're really at in biology that we can now measure this information. And that here at Pepper, we can start to untangle those relationships to truly interpret how to best treat and tackle a disease.

So as you are putting those together, do you look at like different pathways or systems biology or so? Can you talk a little bit about your approach there? Absolutely. So within that dictionary and that functional understanding of these relationships, absolutely includes pathways, includes everything from relationships between kind of connections between proteins, as well as larger pathways in biology, as we call them, that they're related to. So these can be pathways about cell growth, cell death.

about really particular kind of functional avenues of what can happen within a cell. And this is what helps us to relate changes that we see in, say, if we're studying oncology, cancer, we can look and see when there's a change between the cancer and a model for normal tissue, what happens, what pathways are implicated to then address, OK, where do we best intervene with a drug to then reverse or in the case of cancer, really resolve the tumor and remove that?

from the disease. And so, as you look at the drug discovery world right now, how much of a problem is the lack of understanding of the biology of the disease? And do you think that this accounts for the high failure rate of the drugs in development? The lack of understanding of biology is astounding today, I think. We've come a long way, right? I mean, it's something, if you think about in the last 20 years, we've

Daniel Levine (07:54.99)

We sequence the genome, right? But that's the start of it, right? And that's what's so exciting is, you know, we have this instruction set that's been looked at for, you know, for 20 some years now. But that's the instruction set, right? There's a whole new avenue of once those instructions are carried out, what happens, what the implications are. And so while we've gone leaps and bounds in the last 20 years, there's so much more to cover in terms of grounded, truly understanding complex disease.

especially areas such as an oncology, neurodegenerative diseases. There's a lot of changes that are happening at once in that disease state. And these complex diseases are where we're still seeing, you know, failures in the clinic and an inability to actually have a treatment for for a patient that that gap on the patient side is actually for my co -founder and I what's what's really been.

the driving force as to why we're here is because we still have immense gaps clinically for a lot of diseases. And mention neurodegenerative diseases, this has been really a personal experience for both my co -founder and myself. For me, it was my father who had frontal temporal dementia. I'd never heard of the disease until I was in my graduate studies and really quickly learned how much we don't know.

right, on being able to help a patient and having treatments. And for me, I think that was a big wake up call in the aspect of I'd always loved the science and unraveling new patterns, but really seeing fundamentally how much we don't know in answering questions of what's going wrong. And when you have a loved one that's experiencing that, it really brings home all the work that needs to be done still today to truly help patients. OK. And is that something that

drove you and your co -founder to really start this company? Can you talk a bit about that? Absolutely. So the kind of the immediate personal story was really instigated for me during graduate school on seeing the gap for patients, having a family member that there were no treatments available for. Similarly, my co -founder, his grandmother had Alzheimer's disease, so very similar experience that I had there. But

Daniel Levine (10:15.118)

Big picture, we actually, we go back like over 10, 15 years now to having connected my co -founder and I during our undergraduate studies actually at Northwestern University, so just outside Chicago. We were both studying for biomedical engineering degrees there and we're really excited about just making an impact in the healthcare space. We didn't know quite what that meant.

But we both really resonated on that broader goal of knowing there's a lot of new challenges out there and wanting to make that tractable impact. So we kept in touch over the years during each of our studies. Myself, I built out more on the science technical side. He had built out more on the business strategic venture capital side. And we rejoined forces after I had finished my PhD. And he had finished at Harvard Business School and launched.

Yeah, Launch Pepper in 2019 right after all of that. So it was really exciting. Great. Now, you're going after a lot of very tough diseases, the intractable diseases. And you just shared your story about that. But in general, what we see for a lot of these new platform technologies that the companies start out pursuing the well -established targets to establish a proof of concept for the approach. So why are you taking this approach? And

how do you know that that will be successful? Excellent question. So we, I mean, and just kind of big picture context for folks, our own internal pipeline program is going after mick -driven cancers, specifically hepatocellular carcinoma and lymphoma. And so why, I mean, for folks familiar with the mick space, as you pointed out, this is a challenging space. Folks have not been able to directly drug mick itself.

But that's part of why we chose this area is our platform and its ability to look globally at signaling changes at the biology of the disease itself can look for alternative targets, novel targets to treat that disease that are not MEC. So we can avoid looking at that very difficult part that's going to be implicated and look more broadly of what else can we treat or what else can we drug to still treat the disease.

Daniel Levine (12:31.598)

And so for us, it's really the sweet spot of going in in a difficult space that we have an ability to truly characterize this disease effectively. And a little bit of context of our success rate there, we've gone through already validation studies of targets within a range of different cancers and have been seeing a 73 % success rate for in vivo validation. And so for us, that's been first, the huge first proof point of yes, they are working as we go through.

And that's something that we've been really excited to see that success there of as we grow our own pipeline. And as you're looking at these hard to treat diseases, do you find that there are some where you're able to make a headway where there are others that are not based on the pathways or the data that you have? How does that look like?

So I mean, one of the big advantages is our transomic platform and possible proteomic data in the context of functional understanding of a disease that we bring in. This is actually when we kicked off our MIC program, part of what drew a collaborator to work with us. Specifically, we're working with Dean Felscher's lab over at Stanford University, who is a expert in MIC -driven cancers, also has unique model systems to help tease out the biology itself. And so our platform, our ability to see

Again, a global context of data, a functional context of data through looking at again, phosphoproteomic data to understand what proteins are doing. This is in contrast to the status quo where people typically look at how many proteins are there. But frankly, if I don't know what they're doing, why do I care how many of them are there, right? So I want to know functionally what's changing. And that's one of the other big advantages to what we do is looking at that activity context of function. And then lastly, our integrating across.

these multiple data sets via transomics allows us to look at true drivers of disease that goes beyond traditional correlation to look at molecular relationships, bringing in that dictionary I mentioned at the start to truly interpret the unique biology that's happening. Now let's come to data and analytics here. So where does your data come from?

Daniel Levine (14:49.934)

Do you generate your data? Do you use public data sets? Can you talk a bit more about it? Yes. I mean, I guess the short answer is both, all the above. So to dig into that a little bit more, for data on a given disease, we collect, we can work with existing public data. Notably just big picture, this is genomic data through next generation sequencing, likewise transcriptomic data through next generation sequencing, and then proteomic and phosphoproteomic data.

we measure via a technique known as mass spectrometry. So again, this large global amount of data to not, we're not prebiasing by a particular panel or assay. We're asking what's in the sample and looking at it as a whole. And so we can work with public data sets when those do exist. We often do collect our own de novo data sets when possible and that samples are available because there is a few advantages to that. One,

we can work with the most advanced data collection approaches. For those in the Mac spec space, this can be known as data independent analysis, mass spectrometry. A lot of words and letters there, but essentially emphasizing that it's a more comprehensive data set that overcomes historical challenges of missing values that can be seen in these large data sets. So we do often collect our own data. And for our current pipeline program, that's how we've been carrying that out. Other advantages in collecting the novel data is just control.

So looking at, we can have data from all these layers from the same sample. So we're reducing noise in doing that to have a stronger signal to draw from. Also, we can control how quickly these samples are processed. You can imagine if you're looking at a very time -driven functional process of what proteins are doing, you want to be really controlled in how you collect those samples to have confidence that you're, again, not adding noise just due to the sample processing.

And so that's something that we're excited of having built and growing continually such a large transomic database to draw from. In parallel with that transomic data, as I'm calling it there, we have a second set of data that reference data set. And that's drawn really from two different facets. One from public resources of what relationships exist between molecular actors.

Daniel Levine (17:05.358)

We're drawing from known biology. We will bring in known biology that's been established. Alongside that, we have our own internal abilities to expand that by leveraging essentially across different model systems more effectively than has been done historically. And so that gives us the ability to leverage known biology, relate that effectively across different organisms, and have the largest reference data set to then interpret those large scale omic data layers that I've mentioned.

So you're using not only the human omics data, but also from the model. Across the board, precisely. Yes. And this is one of the big advantages of you think of mining known biological relationships. There's a lot of data out there on other organ systems, other model systems. And that's something that we do look across those still taking into account, you know, similarity between species and lack thereof. Right. So we are very cognizant of keeping that.

present in our own analysis, but it is something we want to do the best we can of leveraging what's available to us and curating that effectively. And so what are the different model organisms that you use for this? So we predominantly work with rat and mouse. There is some data that we're now expanding into another mammalian systems as we grow the platform itself. But across the board, that's the predominant.

models that we've worked with, in addition to obviously having a lot of clinical data and human data as well. Okay, great. And so now we have a lot of computing power, so we can get a lot of data, but more data is not necessarily better data. How do you distinguish between more data and the right data? 100%. This is, I could talk about this all day. So many facets. One, first is the quality of samples you're drawing from, right? If you're not starting from a quality sample,

What do you have as your foundation, right? So that's a key facet is ensuring and curating that samples are collected consistently at high quality, quickly. Even when we bring in public data, that's a part of our curation process is review of those methods, because if it's not starting out from a strong footing, we don't want to build something where it isn't, like you said, quality data. A second facet that we look at is really what type of information. Again, I'll mention the benefit of having context of what proteins are functionally doing.

Daniel Levine (19:28.302)

So bringing in fossil proteomic data, this functional protein context of modified proteins really expands our ability to have the right data to ask a question. Again, looking at how many proteins are there is a very different question than what are these proteins doing. If all of those proteins present are turned off, using expression as a proxy for function will give you the wrong answer.

And so that's where looking at information of functional context alongside expression is key to knowing what's actually changing in a disease. So when we're looking at inhibiting a protein, we know what's actually going to happen. Is it relevant to that disease itself? And by expression, you mean the quantity, right? The abundance of the protein. Correct. Correct. Exactly. So protein expression, quantity of protein just amount how much is there in contrast to the benefit of our bringing in.

much of a modified protein is there is that can be used to understand the activity of the proteins in that system. Can you talk a little bit about like when you talk about the switches, right, the phosphorylation. So in proteomics, you can measure protein, but then the protein that's actually doing the function that may need to be phosphorylated is turned on, right? So measuring the quantity of protein is not enough, right? So can you explain that a bit more? Absolutely. So when you have a

you know, protein floating around in a cell. It can do different tasks depending on what molecular switch you add. You can use the state of what all these switches are at. So unmeasuring the amount of phosphorylated protein in this case to understand what put on that switch, the activity of that, what's called a kinase. But you're actually using the ability to measure really the state of these switches. OK, if I know who is on or off and who's responsible for turning them on or off, I can back out that.

protein that actually does that flip switching. And that's then where I can get at that functional context. And it's our reference dictionary that gives us the edge to be able to connect those dots between the proteins themselves and the proteins that turn on and off these switches. So it gives us an ability with this reference data set to do that uniquely. We have a large set to interpret that data from. And it's with this functional context that looks beyond how much is their expression that allows us to truly understand

Daniel Levine (21:52.334)

what's functionally changed in disease. So what proteins are truly driving that disease, not just in how many are there, but as what they're actually doing. And have you done any kind of a assessment about transcriptomics, which is the quantity of mRNA, and then that translates into proteins? So in terms of like really getting the right targets, do you find that the proteomics data usually is more

kind of predictive about the right function versus the transcriptomics or genomics? Do you have a sense of that? So the advantage of our approach is we look across them. So you talked a lot about the functional and protein end because the functional context is one of the really unique facets that we bring to the picture. We do look across these data layers from genomics, transcriptomics, proteomics, and fossil proteomics. And so our advantage comes in is that we look at the interplay between them.

and we look beyond just say agreement, right? A lot of folks today, you might have your transcriptomics department say five proteins are up, your proteomics department say two proteins are up, they look at the overlap and say, that must be the right one to go after. That's a little bit misleading, right? Because there's not always a direct correlation between them. And so our approach actually looks at them in parallel. And so we can look at an integrated analysis that looks across.

And that's where we've seen the most value. In our own target predictions, we've actually benchmarked against standard approaches and seen that we're actually able to perform with our integrative approach four times better at predicting. We like to look at known targets that are approved as our true baseline of goodness, because those are in the clinic and being used by doctors today. So we're four times better at predicting those known targets and standard omic approaches. And we're actually 12 times better than an AI lit search type approach.

So really saw that advantage in our approach in transomics of looking at this integrated process to interpret and understand what's the right target to go after that will be clinically relevant at the end of the day. Okay. You described transomics. Is that the same or similar to multiomics, which is another term that's being used these days? Very good distinction to point out. We do distinguish it from multiomics in that.

Daniel Levine (24:12.864)

Multiomics we see is really a separate analysis of each data layer that looks for an agreement between them. You'll see a lot of places will actually have, again, different departments that are, unfortunately, a little bit siloed in their analyses. And then they look a bit higher up as to where their results agree. Transomics that we do at Pepper is distinct in that we look at all these data types in a singular analysis. There's no separate person working on each one. Each analysis is really owned in that one.

one scope and one project to truly get them in in concept. Okay. And so that's how your approach is then different from the companies that are using multi -omics, which is separate. Correct. Very good question to clarify. Absolutely. Now, in terms of how your platform works, can you talk a bit more about that? So you get the data and then it's doing the analysis. Can you describe that a bit more?

Yeah, absolutely. And I guess I'll use a specific use case of looking at target identifications, since we've been talking about that through our discussion. So we'll take for a given cancer, say, information of a disease state and a model for healthy, so typically a normal adjacent tissue, if looking at clinical samples, to then bring collect data from each patient of these multiple data layers in both of those tissue types, feed it into our platform, which starts out by simply having a, you know,

scaling to compare between, for each patient, looking at the disease to normal. So we're first really asking that question of what's distinct in disease and having that control as baseline. And we're generating signatures via our Transomics platform then based on this full set of data that's been in. We're interpreting via our reference data set and then looking at essentially for targets.

that transomic signature is then compiled into an evidence score that we've optimized our own scoring approach to rank potential targets in the top priority target for resolving disease. And we essentially have a list that we can go down. We found, again, in that prioritization, this is that ranking that's looked at from our sets of prioritized targets, where we've seen that 4x improvement over standard multi -MX and 12x improvement over a standard lit search.

Daniel Levine (26:33.326)

So it's really through this optimization as well that we're able to have for targets, the evidence score really going down what's the right ranking of where that top priority target is. And I'll emphasize again the in vivo validation that we've carried out to also vet those targets, because that's another facet. Has there been an experiment as well that shows those are successful? And that's where to date we've seen that 73 % of the targets we've been validating have been successful, which we've been.

really thrilled by and has also been the foundation to fuel building our own pipeline program as we expand that. Great. Great. You talked about your first couple of targets and then focus on the MIG -driven hepatocellular carcinoma and diffuse large B -cell lymphoma. So you talked about the MIG -driven, kind of like the functional knowledge that you have. So can you...

Talk a little bit about these diseases and why you started there. Absolutely. So we, for our own pipeline program, have gone after diseases that are very complex, have been historically difficult to treat cancers in this case. And we picked oncology for a variety of reasons. One,

It's been seen biologically that these modified switches on proteins are really central to the disease itself. So there was a strong reason also biologically that this approach of our data types that we're bringing in for transomics can start to tease out those differences in an effective way to identify a treatment. Also, our approach has a strong advantage and difficulty to treat diseases because we look globally.

So again, we're looking at a range of different targets we could go after. We aren't honed into only looking at a narrow pathway. We can look outside of those canonical pathways, especially in MIC that's key as it's been difficult to go after that pathway directly. So we can look for non -canonical targets that are unique for treating that disease. And can you talk a little bit about MIC itself, the MIC protein and its function? Yeah, so MIC itself, it's often seen as...

Daniel Levine (28:40.918)

it's been a protein that actually has a full signature with it that's been associated with the change for patients. And it's for that mcProtein directly, the challenge has actually been directly inhibiting that protein and amplification of that protein. And so it's been historically, and this has been going after for quite a few years now, we've had it in terms of a field. And that's been partly where we worked with Dean Felscher's lab at Stanford was to look for novel approaches.

that we're outside of MIC directly and able to see where else can we intervene within the pathways themselves that can look at resolving the disease in a way similar to if we inhibited MIC directly. And this is also emphasized where Dean Felcher's lab has really unique ability and collaboration as they have a model system where they can actually tune turning off MIC. So when we're looking at this model system,

and studying that in our own target prioritization, that's our base case of, okay, what happens when MICK gets turned off? And let's now, through our approach, characterize that and prioritize targets that mimic turning MICK off. And so it's actually able to, for our platform, really comprehensively look at alternatives to directly targeting MICK itself. And so for some of these diseases, do you get enough samples? Because for...

analytics, you need a lot of samples and especially getting clinical samples is a challenge. So how do you solve that? So a couple of notes there. I mean, one, we can work with with model systems. So in, for example, in the work that we're doing with Dean Pilcher's lab is a model system that we're working with. So we're actually able to look at in vitro cells, as well as we can look work in in vivo systems.

We can look at a range of different systems because I know folks might know, we can treat cancer well in mice, but that's not helping patients. And so we do often look at a range of different model systems there. And in oncology for our own programs, there are samples that are available for that. So that's one of the other just scientific reasons that's been a clear place to start. But that's a great question because having enough material.

Daniel Levine (30:59.118)

for fossil proteomics, folks familiar with it, it does require more material than genomics. And so we are also looking at and have worked with samples from other types of material from patients as well. So blood, urine, there are also studies that we're carrying out that's expanding the range of what we can work with. And so that's something that we're excited by the opportunities in the long term, because obviously with patient samples, that's the holy grail to learn from.

for a range of different diseases. For our audience, can you explain a bit what is in vitro and in vivo? Absolutely. So when studying a disease, folks are familiar with these clinical samples you can measure to understand what the genomics are and profile all these different data types I've listed. But when those aren't available, you can do some really creative things of how do you grow similar types of tissues that are grown external to a patient.

So you can have cells that are actually derived from patients that are grown in a little plastic dish, really specific conditions that's known as in vitro. So there are cells that are grown that are, we call a model system for a patient. Obviously it's an individual cell versus tissue, so there will be differences, but it can still give you a really nice test system to learn about a disease and to test out drugs. In vivo, for in life, is looking at...

a model system of using a model organism. So this can be anything from a mouse to a rat that actually looks at a model of the disease that's in a whole organism. I you mentioned earlier, systems biology perspective often really leans on in vivo for a lot of those questions as you see a whole system as well of operating of multiple different tissue types at play to truly understand how a disease progresses. Thank you. So.

You have identified these targets and then they are you're validating those. So what's known about these therapies from the work done to date and how are you taking them to advance into the clinic?

Daniel Levine (33:09.774)

Excellent question. It's something I'm really excited about as we've been right now working on what we're calling our FASTA IND program. So the targets that we're looking at, we've been into two categories. Those that are entirely novel that no one's developed drugs against, as well as those that have assets that have been designed to target them, but failed in previous trials. And so targets that passed safety but didn't work in a disease were actually looking to have drugs that went after those.

repurpose to our specific disease area to get quickly into the clinic. And so that's actually right now we've been carrying out experiments and have seen notably in Patocyclic carcinoma drugs that we're considering in licensing have been performing better than standard of care. And so this has been really exciting to see as we look towards growing our own program and the benefit for both patients as well as for us growing as a company.

is by in -licensing these assets that have already passed safety, our aim is to then go into a trial to first question within patients, ask an efficacy -based question, and to say, does this work in hepatocellular carcinoma? And so from the data we've seen thus far and the assets that we're diligently saying, really excited about the potential of bringing them in -house and moving that forward. Great. And you're also pursuing a dual strategy business model, right? So you're

you're partnering and you're also pursuing your own clinical pathway. So can you talk about that a bit? Absolutely. So I've talked a lot about our own pipeline and our discussion this far. So the other half of that, as you flagged, is partnerships with Pharma. So we have to date successfully completed three partnerships with Pharma, one top five Pharma company in a range of different...

disease areas, so oncology as well as inflammatory and neuro diseases as well, in a range of also different drug modalities. So we've looked at everything from looking at surface targets for antibodies to go against to novel activation modalities within a system, as well as small molecules. And so this is really for us been exciting on multiple fronts. One, in terms of our big picture goal to treat the untreatable, it's more shots on goal, right? We can work with

Daniel Levine (35:30.574)

more companies look at more diseases and really leverage that broader team of the drug discovery industry. Second facet is as a company growing, it's an avenue that we've been thrilled to see early validation in of partnerships of the industry working with us, seeing their enthusiasm and hunger for this technology and bringing in early revenue. That's when something alongside fundraising has been a significant advantage of these partnerships is that facet for us as well. So,

I mean, we have seen that AI in drug discovery has great promise. We have heard about that. There are so far limited successes so far today. Where do you think we are in this evolution and why do you think there have not been greater successes? It's an excellent question. And one, I'd say AI is a phenomenal tool that for an industry with such complex data and challenges that I...

we will see strides in its application. I that the key today is really leveraging what's the right data to learn from in AI machine learning, right? And what is the right information to start with to tease out practically relevant, clinically relevant insights. And so that's where Pepper's approach, we focus on having the right data. AI is a part of our toolkit analytically, but we see having this right data being leveraged.

by those algorithms and approaches is where we will shine, where we'll see those insights in the industry as a whole, right? Starting from quality, the right information will be essential. And that's where I think as the industry continues to apply the approach through both Pepper's technology, other approaches as well, we will see broader strides there. But at the end of the day, it's as you flagged, right? If you're not having the right information coming in, you can't learn from that effectively.

And so Pepperbio has raised $6 .5 million in seed funding in 2023. Not a lot of money given the company's ambitious. So how far will this funding take you and what's the plan for raising additional capital?

Daniel Levine (37:40.462)

100%. So this funding itself is really what's building up the start of our pipeline program of in -licensing and asset. The next fundraiser will really support getting into the clinic and that clinical program itself. And so that's really where we've seen that key inflection point of having an in -house asset that then drives that next large fundraise. And our seed is getting us to that point to trigger that. And one of the advantages to our approach is with Transomics, we can learn so much early.

in the drug discovery process, which also allows us to be very, very capital efficient in that. I think that's something we've also highlighted at Bio last year and in previous conferences as well, is we can learn so much even from early stage data and vitro data as well, which big picture for the industry is also huge because the earlier you can learn the right insights to trigger and higher probability of success, the cheaper it will be and faster it will be to get drugs to market.

Samantha Dale Strasser, co -founder and chief scientific officer of Pepperbio. Thank you very much for your time today.

Pleasure to be here. Thank you so much.

That was interesting. What did you think? It was fascinating to understand the transomics approach that Pepper Bio is taking. I've been familiar with the multiomics approach where you have the genomics, transcriptomics, and proteomics, and then doing the analysis for those and then putting that analysis together. But this is a pretty interesting approach of transomics where...

Daniel Levine (39:18.126)

all of that is considered one analysis instead of separate analysis again and putting together, right? So that's definitely a pretty novel approach and something that I think the industry should evolve to and making those right connections using this as a one combined approach. I think that's that will lead to some interesting insights. Definitely. As you think about diseases like cancer or neurodegenerative disease, we

continue to get deeper and deeper into recognizing the complexity of the biology underlying them. Is this the real promise of AI to unlock that complexity? So we have over 20 ,000 genes and then more proteins like that. And we know now a lot about these, but then there is a lot of interactivity, interplay between the different proteins, how they function.

So there's a lot that we still don't understand even today, right? So for, and many of these diseases that are hard to treat are because we don't understand their disease biology and because that we haven't been able to find the right targets for those, right? So what are the actually the causal genes for those? So taking AI and taking this new approaches where,

where I'm not only doing the experiments on the genes, RNA and proteins and the phosphorylation of those, but then also combining all of that. So these are new methodologies. This is new types of analysis that we're able to do using machine learning and AI. And of course, I believe that that is going to result in uncovering some of the biological insights that were not available before, without this kind of data and this kind of analysis.

Pepper starting with my cancers, these are difficult targets. I've seen a lot of companies that plan to go after difficult diseases, start with more accessible ones as proofs of concepts to their approach. What do you think about Pepper's willingness to start there? So they are focusing on the functional information, right? So getting the different biological information, but then getting that

Daniel Levine (41:40.174)

together, combining that with the functional information is important. And so their focus has been the for the MICK related cancers, right? So they have focused on this one area where they can really go deep into understanding the function of MICK, understanding the pathways of MICK, and then based on that, uncovering the targets around that. So I think that's a it's a very focused area where they can go deep into understanding functions. So I like

the way they've done it and focus on that, right? So that way, yes, I mean, when you think about the hard to treat disease, there are many of those, but then focusing on diseases that are related to a specific function, I think that's what's helped them in driving to get the targets in those specific areas. One of the biggest challenges when you're dealing with large amounts of data is to separate signal from noise.

Samantha talked a bit about how Pepper addresses that. What did you think of Pepper's approach? So I thought there were two dimensions in which she tried to address that. So one is for the same gene that gets translated into RNA and then into protein because they are doing that combinatorial, the combined analysis of genomics, transcriptomics, and proteomics together.

they are able to see what is signal versus noise because if there is really the signal it is going to stand up in some of those, in at least one or two of those omics approaches and then that will, that way if there is noise then they're able to kind of take away that noise because getting that same data from the three different types of omics is going to be, that signal is very much stronger rather than getting it just from one.

So that's one dimension. The other dimension she also mentioned is about the model systems, right? So she talked about the in vivo data. So from mouse or rat data. So because there are functional, what we call orthologs, right? So the proteins, they have the similar functions in humans and mouse and rats. So you can also look at, okay, well, are you getting the signal in the omics data in the humans? Are you also getting that in the model organisms? Because if you're getting that signal,

Daniel Levine (43:57.218)

That helps in really understanding that that is true signal versus noise. And then also, I should talk about in vitro data, which is the human cells that are grown in the labs, right? And so getting the data from there as well, right? So that's this other dimension of using not only just the human clinical samples, but also the model organisms and the model systems. So because they have these very different types of...

systems they're using, like the different types of omics and different types of systems. If there is a signal, that's going to stand out because of this diverse way in diverse ways from which they're collecting the data. You also asked Samantha about the track record of AI and drug discovery to date and why there haven't been greater successes. Samantha talked about having the right data. She said it thoughtfully, but essentially she was saying,

Garbage in, garbage out. Are you confident that drug developers will get this right? We're on a journey here, right? So we're getting better and better. And then it's, so I think we are going to get it right. But I think it's going to be, it's going to be a long journey where, as I mean, I remember as I see the evolution, right? So now we are using many more clinical samples than we were using, let's say 20, 25 years ago.

we are collecting much more omics data at this point than we were doing before. We started collecting, connecting the different types of omics data, right? And also in terms of AI, we're now getting better and better technologies to analyze the data. So there has been improvement in terms of getting more data, getting better data, doing better analytics. So there's improvement in all these different dimensions. So I believe,

we will get better, more and more success, but it's something that we as researchers are learning as time goes. Well, I really enjoyed the conversation and I'm looking forward to our next one. Sounds good. Thank you, Danny.

Daniel Levine (46:09.998)

Thanks again to our sponsor, Agilisium Labs. Life Sciences DNA is a bi -monthly podcast produced by the Levine Media Group with production support from Fullview Media. Be sure to follow us on your preferred podcast platform. Music for this podcast is provided courtesy of the Jonah Levine Collective. We'd love to hear from you. Pop us a note at danny at levinemediagroup .com.

For Life Sciences DNA and Dr. Amar Drawid, I'm Daniel Levine. Thanks for joining us.