First in Human Episode #19 featuring Naheed Kurji

Naheed Kurji, the Founder and CEO of First in Human.

For episode 19 we chat with Naheed Kurji, Co-Founder, President, and CEO of Cyclica.  Find out why the power of AI-based technologies and neo-biotech are the wave of the future. First In Human is a biotech-focused podcast that interviews industry leaders and investors to learn about their journey to in-human clinical trials. Presented by Vial, a tech-enabled CRO, hosted by Simon Burns, CEO & Co-Founder & guest host Co-Founder, Andrew Brackin. Episodes launch weekly on Tuesdays & Thursdays.

For episode 19 we chat with Naheed Kurji, Co-Founder, President, and CEO of Cyclica.  Find out why the power of AI-based technologies and neo-biotech are the wave of the future. First In Human is a biotech-focused podcast that interviews industry leaders and investors to learn about their journey to in-human clinical trials. Presented by Vial, a tech-enabled CRO, hosted by Simon Burns, CEO & Co-Founder & guest host Co-Founder, Andrew Brackin. Episodes launch weekly on Tuesdays & Thursdays.

Andrew Brackin: Hi, I’m Andrew Brackin, co-founder of Vial. Vial is a tech-enabled CRO, offering faster and more efficient trials for biotech companies. Today, I’m here with Naheed Kurji. Hi, Naheed, how are you doing?

Naheed Kurji: I’m doing well, thank you. How are you? 

Andrew Brackin: I’m doing well. And you’re the CEO of Cyclica. Can you tell us what you do at Cyclica, and what Cyclica is?

Naheed Kurji: Yeah, definitely. I am co-founder, president and CEO of Cyclica, a, Canadian headquartered biotech company, but we do have a team in the US, as well as in the UK. Our entire mission as an organization is to unlock the potential of the human proteome. To discover and advance the medicines of tomorrow. We are an “AI” for drug discovery company. Though, I don’t like using that terminology as the basis to explain a company.

 In addition to my role at Cyclica, and on behalf of Cyclica, I also co-founded, the Alliance for Artificial Intelligence in Healthcare, or the AAIH, for which I sit on the board, have been one of the four executive officers since the founding. I was secretary for about three years, and then in 2022, I took over as chair of the AAIH and work alongside a phenomenal group, including Angeli Möller, from Roche, Meredith Brown-Tuttle, former VP, regulatory affairs at Recursion, and Maria Pineda, CEO of Envisagenics. I can talk, from my perspective, a little about the broad application of AI across healthcare, though I do spend most of my time in drug discovery.

Andrew Brackin: Super cool. I read about that organization. It’s really exciting stuff. Why don’t we start at the beginning? Can you explain to our audience the process of drug discovery, and why it takes so long and costs so much money today? Before you guys fix it [laughs].

Naheed Kurji: Absolutely, Andrew. To put some numbers around that, as I’m sure many of your listeners know very well, but, it’s important to continue to accentuate the challenges in both time and cost, different reports will say different numbers, but, on average, it takes 10 to 15 years and costs hundreds of millions to, low billions of dollars to bring a drug to a patient. 

The process there is to identify plausible biological target to design, or discover medicine that fits all the preclinical requirements, to then get that into a viable clinical trial, to get it to the right patient population, and to bring that through both safety, efficacy at different stages to eventually, regulatory bodies, and then to the ultimate patients in the marketplace.

Very difficult. Highly non-trivial. From a technology standpoint, if you look at the past, give or take, 50, 70 years, the business of drug discovery and development has only taken longer and costed more over time. It’s become highly inefficient. In some reports, just that the internal rate of return of drug discovery and development is near 0% or just a little bit north of that.

Over the past, give or take, 25, 30 years, the number of technology companies have entered the drug discovery and development space, and said, “How can we harness the power of new capabilities like cloud computing, access to available datasets, to conduct drug discovery and development in a more efficient way, to shorten timelines, and reduce cost.” In an effort that all of us are aligned to is to do much more for human health.

Again, there’s no magic wand or silver bullet, it is a very complicated space. We could talk a little bit more about that, but, again, to answer your question, it takes a long time because it’s difficult. It takes a long time because biology is highly esoteric. Biology and chemistry together is even more challenging, especially when you have to consider the safety requirements before a drug gets into a patient’s hands, and the efficacy to compete against other potential standards of care.

 Coming from a place of just accepting and understanding that, and then finding the right technologies to have demonstrable value, is a really interesting, very, challenging problem to solve.

Andrew Brackin: I’d love to hear more about that. , Tell us how Cyclica uses AI, and how does that approach differ from traditional target-based approaches?

Naheed Kurji: Before answering a question, I start from a place of context at all times. As a student of the history of drug discovery and development, I evaluated, and spent time with a lot of folks, read a lot of papers to understand how things were done to get to a place where the IRR was close to zero. What are the potential opportunities to supercharge that in a different way? Not just incremental, changing a solution for the same problem, but how do redefine the problem?

 If we go all the way back, the reductionist approach to drug discovery is what is the prevailing strategy still today, even though there is new solutions? That reductionist strategy was, for a given disease, to identify a biological driver of that disease, oftentimes a protein that was not functioning correctly, and to identify, then, a specific compound, chemistry to interact with that protein to have the right biological effect.

For the more lay audience, you could think of the biological protein target as the lock, and the chemistry as the key. For a long time, the paradigm was: let’s identify one key for one lock went to a specific disease. One chemistry for one protein for a specific disease. That reductionist strategy is what the the industry continued to innovate around, both empirically, through wet lab chemistry, and then through computational techniques in the late ’80s and early ’90s.

The first wave of computational approaches to drug discovery was to take what others would do, in PieTechs with chemistry and biology, and to simulate that in a more computational framework. The first companies introduced docking-based technologies to simulate physically how a compound molecule would interact with a biological target. 

They did this through various, really elegant techniques, like molecular modeling, through energy perturbation, et cetera. What started to come out of the mid to late ’90s, early 2000s, was a supercharged way of doing what was done empirically. Now, screening millions of molecules rapidly against a biological target, in a more digital way.

That’s been the prevailing computational strategy, or that was until around 2010 to 2015 timeframe. Around 2002 to 2010, with the Human Genome Project and a wealth of data that started to emerge in the landscape, a number of companies started to introduce themselves to drug discovery as big data prior to there being language around AI for drug discovery, the space was called big data and predictive analytics. 

There was a seminal paper written by McKinsey and company titled How Big Data Can Revolutionize the Pharmaceutical Industry. That was published in about 2013. On the back of that, give or take, a dozen companies started to emerge in the big data predictive analytics space.

A vast majority of those companies still focus on that same problem which was one key for one lock for a specific disease, one molecule for a given target for a specific disease. That, fundamentally, is not the problem that Cyclica set out to solve. 

Whereas everybody is focused on building the next best docking technology. To screen millions to billions of molecules, and force fitting them into a specific target, or bringing kind of new AI techniques, like the ligand-based models, or quantitative structure-activity relationship models, to identify molecules that look like active molecules for a specific target, we flipped the problem on its head, and we said, “Let’s build one model for the entire proteome,” the collection of all protein targets, so that we could harness the strength of a molecule’s polypharmacology, the idea that it interacts with dozens of targets, not just one.

Andrew Brackin: It’s incredibly interesting. Can you tell us how the Human Genome Project impacted your work at Cyclica? It’s obviously been, a project that, has been many years in development, right? 

Naheed Kurji: It’s impacted us, the entire industry, and therefore companies like us measurably on a few dimensions. Number one: mindset. Data is very important. The access of that data in the development of representative models that can be applied to the application of drug discovery and development is critical. 

The first massive wave of that data came post-Human Genome Project. There’s a lot of data that is required for companies like us. There’s protein structure information, there’s drug-target interaction data, there’s genetic and genomic data. A lot of that last piece of data, genetic, genomic data, emerged, and continues to grow exponentially as that data becomes more available. That would not have been possible if it was not for the work that was put into the Human Genome Project, and the first wave of that data coming out in the early 2000s.

What then is available to us? Well, more protein [00:10:00] structure information. More ligand protein data is now available. More insights into personalized medicine is now questions that can be asked. And the challenge with genetic and genomic data is that approximately 90-94% of genetic and genomic data come from patient populations that don’t necessarily look like me. Quite frankly, Andrew, they look like yourself. Affluent, Caucasian males disproportionately represent, genetic and genomic data.

In the world then, coming back to my comment of personalized medicine, how can we truly talk about personalized medicine where the data that is being used to train models that are then used in drug discovery do not come from a diverse patient population pool? Now, there’s a lot of work to go and fix that, but still, up to this point in time, it’s very homogenous. 

A lot of the effort then, at Cyclica, as well as the Alliance for Artificial Intelligence and Healthcare is access to fair and representative data of patient populations. Because we hear the saying, “Garbage in, garbage out,” for AI models. Fine, that’s catchy, but it’s a bit cliché. It’s now biased data, or under-representative data leads to biased models, which lead to biased outcomes. And I think that’s garbage.

Andrew Brackin: I’d love to learn about some of the strategies you’re using to solve that problem. What are some of the, early tactics you’ve looked at?

Naheed Kurji: If I come back to the Cyclica problem. Then there’s the AAIH stuff, which flow hand in hand, but I’ll talk for Cyclica more micro business day to day, and then stuff at AAIH.

At Cyclica, the question that we asked ourselves, and the problem that we set out to solve was not how to find the next best molecule for one protein target. Why? Well, even the most highly potent molecule for a specific biological target, when placed into a complex biological system like an animal or a human, will interact with dozens to hundreds of potential biological targets. 

Now, some of those unanticipated off-targets can have deleterious effects. It could be linked to targets or it can be interacting with targets linked to toxicity, like hepatotoxicity, or a liver toxicity, or cardiotoxicity. Or they could be interacting with targets that are linked to other disease opportunities. An indication expansion or reimagining new use cases for existing assets can be pointed some place else where efficacy may be higher than the intended target.

The idea of polypharmacology to us was critically important. A given small molecule interacts with, dozens, up to about 300 targets based on literature and our evaluation. We found it, out of the gate, a disservice to build a technology that was simply faster, and maybe less expensive to do single-target drug discovery when what about everything else? The innovation on the problem statement is what really created Cyclica. That’s why we created one model for the entire proteome.

The solution that we brought to that problem initially, and when it combined both the history to today, we brought a docking-based approach to proteome-wide screening. Where we created an innovative strategy to take a small molecule, screen it against the entire proteome, but through a physics-based docking strategy. What we learned from that is that docking is slow, it’s costly, and the predictive power is not all that inspired. The signal to noise is challenging to interpret.

We knew the problem statement was the right problem statement, but our solution set was the wrong solution set. That led us to innovate the technology that we brought to the market called MatchMaker. A deep learning framework that is trained on multiple types of data, but in particular, protein structure data, and drug-target interaction data. 

With the advent of new protein structure technologies like AlphaFold from DeepMind, and the work that they introduced and made openly available to the scientific community in July, 2021, Cyclica, to the best of our knowledge, was the first company to address all of that protein structure information from AlphaFold and AlphaFold 2 directly into MatchMaker.

 We were able to build the single largest, most robust one model, generalizable across the entire proteome. That now allows us to do drug discovery not just for high-data targets that are compatible to physics-based modeling, or AI-based technologies where there’s a lot of data, but we’ve built a remarkable approach to transfer that learning from high-data targets into low-data targets that are incompatible to physics-based technologies or AI technologies. That are generally recalcitrant to even the empirical strategies. That’s where we have an impact as a company.

To your question, polypharmacology is important. One model for the entire proteome. Bring in a [00:15:00] strategy of drug discovery to go after high-data targets and low-data targets. Those low-data targets are linked to diseases that nobody’s working on, or is struggling to work on today.

Andrew Brackin: In 2019, you announced the joint venture with ATAI Life Sciences to explore how psychedelics can help patients dealing with mental health disorders. I’d love to learn more about this company. Also, just other collaborations you’re exploring. You’ve told us about your technology, but would love to learn how it’s manifesting into new drugs and companies that will help patients.

Naheed Kurji: With the access to, effectively, the entire proteome, we are a disease-agnostic company from a technology point of view. We’ve done work in infectious disease, cardiovascular, ophthalmics. A lot of work in oncology, and inflammatory-based diseases. We started to see where the technology was really strong, and where there were some limitations. That’s all part of the journey of building. We can’t call ourselves an AI company if we’re not willing to learn from where the technology doesn’t work as well as where it works.

We started to get more informed on directionality, and the demand of applicability of our technology. Over the past three years, we’ve narrowed in on a biology thesis of immunoinflammatory-based diseases. A focus on CNS, oncology, and on autoimmune disease. Autoimmune disease is one of the reasons why I helped co-found Cyclica. Family suffers from autoimmune disease, and there’s stories that I’ve told about family history in the medical system, where things just did not work out well. 

As a result, autoimmune disease is something that we will continue to focus on as a business. But if you think about the portfolio strategy as CNS and oncology are primarily the two, and then autoimmune disease is tertiary.

 Within CNS and oncology, those are two big areas, right? There’s brain disease, pain, oncology. A huge space. The reason why is generally twofold. For CNS-based diseases, we noticed that a lot of literature has been written about the fact that most brain diseases and diseases of the central nervous system are highly complex. 

Single-target drug discovery strategy is incompatible to actually measurable or demonstrable health outcomes. A lot of people and papers have been written about neurodegenerative diseases, and taking a pathway view, and a panel-based view to drug discovery that there is multiple targets implicated in the degeneration of the brain. Going after just one target is probably inconducive. 

These polygenic based diseases, oftentimes are diseases of the brain and the central nervous system, are so compatible to our system. By accessing the entire proteome, we’re not just going after one target, we can go after multiple targets. We can take a panel-based view to drug discovery. 

We’ve started working in both neurodegenerative and neuropsychiatric back in about 2018/ 2019. That’s when I got to know the leadership team at ATAI at a conference in London. We got together and were talking about our interests in neuropsychiatric diseases. Specifically, mental health. We talked about depression, autism, and bipolar disorder. We were just reminiscing about family history in those areas.

We said, “Why don’t we start a company?” That’s what then led to ATAI and Cyclica creating EntheogeniX Biosciences. A company, I believe, is one of the leaders in the space of understanding how psychedelics work. Then, harnessing that understanding to discover new medicines that will recapitulate the benefits of psychedelics, but in a more NCE regulatory path. We’re really excited about the work that we’re doing right now. 

Andrew Brackin: Incredibly exciting. I know patients in that field have such limited options. It’s really amazing stuff. 

At Cyclica, you’re developing a large pipeline of assets. I’ve listened to some of your interviews. You mentioned you’re moving towards running more programs yourselves. What does the future of clinical trials need to look like to support that? Hopefully many programs going into clinical trials. Tell us what would you need to exist to support that world, that hopefully will, come into fruition soon?

Naheed Kurji: That’s the best question you could ask, and one that I’ve been obsessively thinking about. Again, I start from a place of context. For your listeners, I’ll answer the question, but let me just back into the answer.

 The traditional business of biotech is: have a platform, direct that platform, or one or a couple specific programs. Invest all of your energy to advancing that as far as possible. Ideally, into a phase one or a phase two. Get it to a specific inflection point. Partner out to a biopharma company who have the infrastructure to take that to patients. That’s the business of biotech.

 With the power of AI-based technologies, where you can move much faster, cost is lower, and throughput is higher, you now have orders of magnitude, more efficiency, to bring to drug discovery. If you have a technology that can work that fast, and that much more efficiently, does it not behoove you to do much [00:20:00] more with it? 

It’s like taking a Ferrari, and then just leaving it on the street of a city, and being like, “Oh, should I not take this on the free road?” The question that we then asked ourselves was, with the power of technology and a new problem statement, can we reframe the business model of drug discovery?

 One of the ways we got into the business of the discovery of medicines instead of what, historically, was our business model of servicing pharma with our platform and trying to generate revenue, which was the very early iteration of Cyclica up until about 2018. Since then, we turned the platform in on itself, and we became a, neo-biotech. A new-age biotech company, powered by technology in a radically innovative business model.

 We’ve raised just over $40 million of capital to date. We have a portfolio of assets that are north of 60 [million], with multiple and late-stage preclinical and ID-enabled studies. Some in late-stage lead optimization. Hit to lead, hit discovery, and then a wealth of early-stage pipeline that we haven’t even yet worked on.

The question is: how are we able to do so much with so little? One of the ways was through partnerships where we didn’t have to own the entire value chain. We knew that there were a number of biotech companies and academic institutions who had extra casing capabilities that we did not have, that were directly adjacent to our core expertise. We combined forces, just like with ATAI and EntheogeniX.

 By virtue of that, we were willing to split the economic upside in the intellectual property, proportionally. They would cover their costs. They would get their upside. We could cover our costs. We would get our upside. Again, the question we asked ourselves: would we rather own 100% of one or two programs, or 50-plus % of hundreds of programs? The first iteration of Cyclica to build our portfolio was to do it in partnership.

As we started to learn more, domain of applicability, strengths, limitation, we raised more capital, of course, we want to take more control of our destiny. We want the decision making to reside with our drug discovery team. Over the past year, we brought on Mike Palovich, our chief science officer and head of drug discovery, 24+ years at GSK before Cyclica. With him now at the helm of our drug discovery team, we said, “Let’s go deep,” and actually, instead of relying on partnerships 100%, let’s shift the allocation. Let that teeter totter, and make it more 50-50. And then, over time, more internal, less reliant on partnerships.

There’s an economic model that drives our business model. How do we keep our cost of capital down, increase our probabilities of success, and then, over time, make sure that we are a financially viable company with the eventual commercialization of those assets. 

And the last thing I’ll say, Andrew, is, when it comes to commercialization, we don’t have an aspiration to run a clinical trial. We know the lane in which we swim. We do drug discovery really, really well, and we’re only gonna get better.

We believe that the market is looking for companies to create, at-scale, early-stage pipelines to get to the preclinical stage, at which point, pharma companies have dried-up pipeline, early stage. Can we create a marketplace, what we call a molecule marketplace so that the wave of drug discovery assets of the future will come through our pipeline, commercialize to pharma companies, and have a revolving door? That’s what we’re really excited about, and that’s what we’re working on.

Andrew Brackin: I appreciate the time today. Thanks so much, Naheed. I wish you well on this amazing journey.

Naheed Kurji: Thank you, Andrew. And I appreciate you inviting me to this conversation. 

Andrew Brackin: Thanks so much.

Connect with us.

Interested in receiving a proposal from Vial? Leave us a message and some of your contact info and we’ll be in touch with you shortly.

Name(Required)
By submitting, you are agreeing to our terms and privacy policy
This field is for validation purposes and should be left unchanged.

Contact Us

Name(Required)
By submitting, you are agreeing to our terms and privacy policy
This field is for validation purposes and should be left unchanged.