Toronto-based Cyclica President and CEO Naheed Kurji acknowledges that artificial intelligence (AI) is a transformative technology, but he contends it is not the “silver bullet” for drug discovery and development.

Instead, he says that AI together with cloud-based computing could serve as a catalyst for a new approach to drug development. Kurji emphasizes that it is important to create “a virtual drug discovery ecosystem where a number of companies who are expert in their space come together and present a more holistic solution than any individual one could do itself because there is no one silver bullet to this problem. The market is so big and there are so many issues, one company can’t do it alone.”

Kurji leads a five-year-old company that has developed and validated a cloud-based platform, called Ligand Express, which uses biophysics, bioinformatics and AI to help pharmaceutical companies navigate the drug discovery pipeline by assessing the safety and efficacy of drugs. The integrated platform enables companies to screen potential small-molecule drugs against repositories of structurally-characterized proteins or ‘proteomes’ to identify significant protein targets. The platform then leverages AI to determine the biological relevance of these targets, and systems biology data to link this information to particular biological pathways or diseases.

Kurji says Cyclica’s platform, broadly launched in November 2017 already is being used by some of the top 50 pharma companies globally. “AI is certainly important,” he adds. “We leverage AI, machine learning and deep learning, but unlike other companies, we also use computation to model the detailed mechanism of drug action. We believe that a comprehensive approach to drug discovery requires a combined approach, and that AI is most powerful when it builds on biophysics and chemoinformatics approaches.”

Kurji received his MBA from the University of Toronto, Rotman School of Management. Prior to joining Cyclica, he worked as an associate director at the Bank of Montreal in Calgary where he was integral in managing a large portfolio of clients across diversified industries.

WuXi AppTec Communications’ interview with Kurji is part of an exclusive series spotlighting the inside perspectives of thought leaders on topics shaping the future of new medicines.

WuXi: How is your company applying artificial intelligence (AI) to drug development and how does your approach differ from other companies?

Naheed Kurji: AI may be the buzzword of today, but it is not a new term.  It was first described by Alan Turing in the 1950s. The second thing to know is that computational drug discovery has a history that goes back to the 1970s. So really what we’re seeing today with the hype and buzz is some exciting tangible value driven by 60-plus years of work in this space.

Over the past few years, driven by the increase in computational power and availability of data, numerous companies have emerged to tackle the drug discovery problem by applying AI to large databases of known experimental drug binding information or other forms of relevant data.

In our view, it’s undeniable that this approach will play an important role in the future of drug discovery, particularly increasing productivity and R&D innovation. But while AI is certainly useful, we firmly believe that it’s only one piece of the drug discovery puzzle. Another very important piece is the detailed biophysical interaction between a drug and its target. The nexus between AI and biophysics is critical to bringing better medicines to market at lower cost. Individually, they are not as strong as they are when they are combined. We believe that a comprehensive approach to drug discovery requires a combined approach, and that AI is most powerful when it augments biophysics and chemoinformatics approaches. So, at our core, we are a biophysics company that leverages AI to augment biophysical findings.

WuXi: What is a specific example of how you apply your technology?

Naheed Kurji: At Cyclica we have a cloud-based platform, called Ligand Express. It is used by scientists in pharmaceutical companies. We’re not doing the work for them – they now have access to our integrated technology platform to augment their scientific discovery efforts. They are the experts. Our vision with Ligand Express, and the continued innovation behind the scenes, is for it to be an integral utility platform in the pharma R&D value chain by augmenting, not replacing, the valuable efforts ongoing in pharma. Armed with an integrated workflow, a medicinal chemist, biochemist, or traditional bench scientist can make more efficient decisions.

Ligand Express is broken into three individual steps. The first step, for any given small molecule drug – whether its hypothetical, preclinical, clinical or FDA approved – Ligand Express takes the 3D structure of that and through computational biophysics uncovers all the protein targets that interact with that small molecule. And what’s important is that step one – with a proteome screening where we uncover all the targets – is based on biophysics. We are able to uncover at a molecular level how a drug interacts with a target.

You can think of it this way: The platform provides a unique panoramic view of a small-molecule, by identifying on- and off-target interactions that may be expected, as well as those that are unanticipated. Importantly, because we are taking a biophysical approach, the system moves beyond canonical binding/active sites and also considers novel interactions and allosteric sites.

Machine learning and AI are not taking that approach. AI relies on known information to make predictions whereas biophysics is discovering novel information.

In step one, you have a bunch of proteins that interact with a small molecule. In step two, you make sense of that data through AI and machine learning algorithms by making a prediction on modulatory effect. Step three leverages bioinformatics and AI to uncover the pathways in the system, links the data to diseases, evaluates drug to drug interactions and protein to protein interactions, and presents the findings as a drug-protein interactome.

When you look at the fully integrated platform, in essence we use molecular modeling to uncover novel information, and then leverage AI to link that with existing information. By understanding how a small-molecule drug will interact with all known proteins, Ligand Express provides value in elucidating mechanism-of-action (e.g., phenotypic screening deconvolution), prioritizing lead candidates, understanding side effects, as well as determining new uses for existing drugs, or what is called drug repositioning.

WuXi: What do you see as the major benefits of AI? Will we produce drugs faster? Will it reduce the trial and error that occurs in clinical trials and drug development?

Naheed Kurji: Our view is that with AI augmenting biophysical approaches, it is now possible for the first time to understand many if not a vast majority of the steps in development of a given drug computationally. So we can reduce the trial and error that happens in the lab. We can provide new information to augment lab testing by providing new, legitimate hypotheses through in silico computational modeling. This is possible today, whereas five or 10 years ago it just wasn’t possible. AI won’t replace scientists. It will present them with new information so they can do their job more efficiently, more productively. Machine learning is a way to process the information, ferret out the important parts, and present a coherent picture to the researchers, who will then use their human judgment to make final calls.

The major benefits of AI will be to dramatically increase the rate of innovation and ability to expose new insights and trends. That’s critically important. The amount of data and literature on the biology of disease is now so voluminous, that no scientist can read and comprehend it all. With AI, it becomes possible to sift through this information, present what is relevant to the scientist, and arrive at intelligent inferences and predictions much more efficiently than any unaided scientist could. In drug development, AI can help design better clinical trials and stratify different patient cohorts in a more organized fashion. AI will be a major influence in the application of personalized medicine and pharmacogenomics. The big theme is that ultimately AI algorithms will augment drug discovery and development, and lead to better outcomes faster for patients.

WuXi: What are some other limitations in applying AI?

Naheed Kurji: AI is reliant on three things: high quality data, high quantity data and data that are relevant to the research question being asked. These data must already be known. AI algorithms are parsing through known information and making inferences and predictions based on that known information. So AI is great at distilling value from large amounts of information. But as Andrew Ng of Stanford University said, ,Data is the rocket fuel of the AI engine.’ I love that quote, because by extension, when the information is sparse, AI by itself struggles to provide value. This can be seen in our industry in rare orphan diseases that are not well studied or therapeutic areas that have long been neglected. To predict previously unknown binding sites, computational modeling at the molecular level is needed. Machine learning, AI, deep learning – those algorithms will not suffice because they are reliant on known information. This is where biophysical modeling is helpful, by providing new information on targets or pathways that are not well studied or lack high quality data. I don’t believe this is spoken about enough. It’s demystifying AI.

AI and biophysical modeling are indisputably complementary. AI sifts through large amounts of information. Biophysical modeling creates information de novo, through in silico experiments. Another limitation of AI, besides data volume, is data quality. Model quality data will drive AI’s innovative rate. You can’t just throw a plethora of data into an AI algorithm, that’s kind of useless. Garbage in garbage out – we’ve heard that adage time and time again. Model quality data is really important. It’s going to cause organizations and researchers to approach data in a more meaningful and focused way, where the outcome is model ready and very reusable.

Lastly, we don’t know what we don’t know, which means humans still will have to perform basic research to add to the current understanding and knowledge. There is still so much knowledge about biology that is unknown. We will still require biophysical modeling to uncover binding sites and leverage AI to make more sense of that more expediently.

WuXi: How will AI evolve over the next five to 10 years?

Naheed Kurji: Generally speaking, most people approach this question with more technology, with more science, and more data. That’s important, but I believe answering that question with only that approach is insufficient. We need to think about the psychology. At the end of the day, I consider myself a behaviorist, or at least someone who is passionate about understanding human behavior. I run a thriving, budding technology company in the bioinformatics space, but I’m addicted to the idea of evolving people’s behavior to better utilize technology to make decisions. I believe that for AI to best be adopted in any industry, you have to think about the psychology of the people who are using it, and how they are going to adopt it. That’s really important.

In drug discovery, it requires a patient strategy and a long term view. We can’t throw technology down the throats of scientists, especially of bench scientists who spent 10-plus years of their lives getting a PhD in chemistry or biochemistry. We’re not going to succeed by approaching them with the attitude that “you’re going to have use it because it’s the next best thing, trust us”. We have to show them how they can be better scientists by adopting and leveraging our technology, and that comes from up front validation work, evaluation through testing, and integration that augments, not disrupts existing workflow.

So I think the next five-to-10 years, first and foremost, are going to be a learning exercise from the computer scientists and technologists on how to interact with the scientists, so they together are stronger than each individual by themselves.

Coming back to the technology side of things, there’s so much data out there, but historically, and even now to an extent, a lot is not accessible. Far too often in pharma the data are kept in silos or hoarded by particular research groups or within research groups by particular scientists. They have this “they own that data mentality.” That’s a limitation because for AI to be effective they need access to that data.

We’re seeing this change. In the past 24-to-36 months, there have been really salient changes happening in the industry. There is this shift opening up with data, and sharing it within institutions. We’re even seeing some efforts where data are being shared openly across the ecosystem. The Structural Genomics Consortium (SGC) is a great example. It’s a precompetitive environment where pharma companies have come together to support the SGC’s mandate of an open concept in structural biology. This is now open data to the marketplace. Gaining access to more data will, of course, be a determining factor to the success of AI; and we will see more of that, whether it’s genomic data, proteomic data, drug binding data, or access to electronic medical records. In the next five to 10 years we will figure all this out.

WuXi: How soon will AI become an integral part of pharma and big biotech R&D and can they survive without it?

Naheed Kuji: The answer to the first part of that question is ‘now.’ I think 2018 is a monumental year for pharma and technology companies. You’re seeing big pharma companies include on their expense lines investment in AI. Almost across the board you’re having big pharma companies attend AI conferences. You’re seeing them invest in human capital to bring AI expertise in house. Three or four years ago, it was a conversation, and there was a lot of skepticism and reluctance in the marketplace. There was a lack of engagement – true dialogue – and pilot projects were few and far between. But if the trends that we’re seeing can be extrapolated across the space, the answer to the first part of your question is ‘today.’

The second part of your question, can pharma survive without AI? Looking forward 20 years, I think the most impactful and valuable pharma companies will be those that were early adopters of novel technologies and those that thought about the future differently from convention today. So I think they can survive, but there will be a shift in the pecking order, not unlike what we have seen in other big industries. It’s not going to be that over the next few years you will see a pharma company that doesn’t adopt AI wilt and die. But over the long-term, if they are not willing to extend themselves and try these new innovative efforts and collaborate with these companies to bring together their respective knowledge and expertise, they risk being disrupted, or as one vice president of chemistry at a big pharma company once told me, ‘If we don’t try this, we will be left in the dust.’

WuXi: What kinds of partnerships are important to your development?

Naheed Kurji: Partnership and collaboration are the DNA of Cyclica. We are good at what we do. We know what we know, but we accept that there is a lot more to learn. Collaboration with academic groups is important because that’s where much of the basic research is done. Collaboration with the biotech industry is important to understand where some of the early innovation is going. And collaborations and partnerships in co-designing new technologies, or refinements of our existing platforms with big pharma, who are the users of our technology, is critically important. So collaboration is key, and if you look at Ligand Express as it is today, it is a product of market feedback that we received through early collaborative efforts, particularly in the academic and biotech space. Looking forward, for Cyclica or any other technology company to have a big impact on the market, it will have to be more open to collaboration and partnership.

As an ecosystem, in terms of innovation in life sciences, too often it’s a cut-throat industry where people don’t talk. They don’t share ideas and they tend not to collaborate well. There are exceptions to this rule, of course, but in general I believe that holds true. This occurs in more than just business, this occurs in contribution to science, in contribution to medicine, and in contribution to human health.

In the world in which we operate, our goal is to catalyze a virtual drug discovery ecosystem where a number of companies who are experts in their space come together and present a more holistic solution than any individual one could do itself because there is no one silver bullet to this problem. The market is so big and there are so many issues, one company can’t do it alone.

These experts can be from bold and innovative pharma companies, forward thinking venture capital firms that are propelling new innovation forward, global consulting firms that are disseminating to the market thought leadership on hot button topic, or even existing technology providers that are building their presence in the life sciences. Over the past 12-to-18 months, we have been talking openly about how we can bring together these experts to shape the narrative about the future, and over the next few months we will take big steps in bringing these partnerships together to take Cyclica to the next level.

WuXi: How soon will patients see the benefits of this new technology?

Naheed Kurji: Patients, whether they know it or not, are actually being impacted by AI today. There’s the whole personalized medicine wave, genomic profiling, preventative medicine, and the availability of information accessible to our smart phones about oneself. That’s all driven by AI algorithms. That’s today. From a health care perspective patients already are being impacted.

You’re seeing AI in the whole field of drug repurposing and repositioning, and leveraging in silico modeling. AI is helping to extend applications of marketed drugs for new indications. It’s being used to look for potential contradictions between certain drugs. If patients have a genetic make-up that is detrimental for taking drug x and drug y together, the doctor can have that information available at the click of a button. That’s pretty close, if not there already in prototype stage.

WuXi: What are the implications of using AI for the reduction in cost of drug design and development?

Naheed Kurji: Undeniably, there will be a cost saving element in the near future by applying AI to specific research questions that otherwise would be onerous or lengthy to do solely by traditional approaches. We know the numbers – 15 years and $2.5 billion to develop a new drug. Eroom’s Law, the reverse of Moore’s Law, applies to pharma – significant productivity decline, more money spent every year and fewer drugs coming to the market. The market is looking for cost savings. AI presents a significant opportunity to compress those 12-to-15 years of development into seven to 10 years because you can take out a lot of the upfront trial and error in drug discovery. From hit to lead, you can compress all that to help scientists take more steps in the right direction and fewer steps in the wrong direction through technology.

From a drug discovery and development standpoint, the impact is diluted a little. It’s not heavily spoken about because it has to go through so many different waves of iteration of lab testing and clinical trials. But a lot of that stuff is being compressed, and productivity gains will soon be realized. 2018 is an important year; and going into 2019, I wouldn’t be surprised to see some successes from companies like Recursion Pharma or Deep Genomics. You’re already seeing AI companies like Exscientia, BenevolentAI and others win sizable deals with pharma.

Our platform is a risk adjustment tool, helping pharma companies identify novel targets expediently. For example, phenotypic screening is an approach in which pharma has invested heavily, but uncovering the mechanism of action in that phenotypic screening is lengthy and onerous. It can take months, if not years, and cost millions of dollars to uncover the target that explains the biological observation seen in a phenotypic screen. With our platform you can hone in on biological targets computationally within days. You’re seeing tangible benefits, from years to days, and millions of dollars to hundreds of thousands of dollars or lower. And so there is some significant cost saving there.

We continue to generate more commercial traffic with big pharma, biotech, and research institutes, while we develop innovative next generation complementary technologies that extend our value in the marketplace. We are energized by the traction we’ve received, especially in the early part of 2018 with big pharma, and will continue to focus steadfastly on enhancing scientific discovery so that together we can more efficiently get patients the best medicines. Overall, we believe that AI, when applied effectively, specifically by augmenting biophysical simulations, can have a demonstrable and significant impact in the future of drug discovery.