In this Q&A with Pharmaceutical Executive®, Andrew Hopkins, founder and CEO of Exscientia, reveals how artificial intelligence (AI) is currently being utilized in the pharma industry, predictions for where AI can be implemented in the future, and what pharma fears most about AI.
As artificial intelligence (AI) hovers between a “buzz phrase” and marveled reality, many point to the potential the technology has in the pharma industry. While some criticize the rate at which pharma uptakes new technologies and implements them into their processes, the industry must also strike a fine balance with moving cautiously when it comes to changes due to the nature of the industry and the potential impact on patients those changes can bring. But it can’t be denied that there is a plethora of data in every part of the process of drug creation and distribution that needs to be more efficiently managed and utilized—from preclinical and R&D to manufacturing to commercialization and distribution. But perhaps one of the greatest appeals of AI is the potential it has for innovation.
Andrew Hopkins, founder and CEO of Exscientia, an AI-driven precision medicine company founded in 2012, shares in this Q&A with Pharmaceutical Executive® how AI is currently being utilized in the pharma industry, predictions for where AI can be implemented in the future, and what pharma fears most about AI.
Hopkins: Until recently, large parts of the pharma industry resisted AI and machine learning, as these technologies were seen as disruptive to conventional ways of discovering and developing drugs. In fact, when Exscientia was founded more than a decade ago, we had to swim against the tide to make the case for utilizing AI in drug discovery with potential partners.
But the dam has broken. Most companies are now embracing AI in some capacity, using tech applications to inform some aspect or other of their R&D work. That said, many are applying AI as a piecemeal or “add-on” strategy instead of taking an end-to-end approach that leverages today’s available computational power to proactively select the right drug targets, design the right molecules, collect the right data, and select the right patients to benefit. While some biopharma powerhouses like Sanofi or Bristol Myers Squibb are beginning to integrate AI more fully in their drug R&D process, we have yet to see the industry as a whole fully embrace the end-to-end AI approach that Exscientia believes is required to transform medicine discovery and delivery of the future.
Hopkins: We’ve only begun to scratch the surface of the potential of AI to help develop new discovery methods that may enable us to identify new therapies more likely to work. At Exscientia, we believe that all new drugs will be designed and developed with the help of AI within this decade. Secondly, I predict AI will increasingly be used to guide the identification of the right treatment for individual patients.
Our industry has largely been shaped through failure, using a risky and costly process of trial and error. This traditional model relies on a small number of clinical successes to subsidize our many R&D failures. And so many drug candidates typically fail in the most expensive, latest stages of clinical development after years of study and investment.
AI will—and already has—help us screen, design, and select high-quality drug candidates using the innovative technologies and computational power available today. This is beginning to significantly reduce both the time and the number of synthesized compounds needed to identify viable new compounds for clinical development. It helps us reduce the industry’s high historic failure rate by advancing only candidates with high probability of clinical success and those we can learn from most. Exscientia has already shown this with six AI-designed candidates brought to the clinic to date. In doing this, we averaged about one year from hit identification to clinical development candidate, compared to the industry average of four to five years.
Hopkins: I think many people still look at AI in biopharma simply as a speed and efficiency play—a way to accelerate drug development and, thereby, lower the cost. But the beauty of end-to-end AI is that integrating AI with automation also offers a path to increasing the quality of therapies and ensuring the right patients may benefit from innovative future drugs. Exscientia’s AI platform aims to bring more effective medicines to patients by analyzing thousands of different parameters in parallel, working in a computational space far beyond the ability of the human brain to process.
Another assumption may be that the animal model is the gold standard for preclinical testing. We believe animal models are an important research tool but do not sufficiently represent human biology. Exscientia’s approach tests promising drug candidates directly on live human tissue samples to record drug effects at the single-cell level. In the EXALT-1 study, we demonstrated for the first time that an AI-powered precision medicine platform can propose which therapy could be most effective for patients with late-stage blood cancer. Patients treated with drugs selected by our AI platform showed a significantly higher overall response rate and responded significantly longer than on their prior line of therapy. EXALT-1 marked the first time a functional precision oncology platform was shown to improve patient outcomes in a prospective interventional clinical study. Using AI, we can look at real patient tissue samples to better understand which drugs are likely to benefit the individual patient.
Hopkins: AI-powered drug discovery and development is disruptive to the traditional pharma business model and may pose challenges to that established model—how R&D has been organized and run for decades.
That said, I think we’re making enormous progress in demonstrating our value proposition and overcoming those fears. And even with end-to-end AI utilization, human beings are still in control. We’ve got brilliant scientists collaborating with our tech experts at the intersection of AI, chemistry, physics, and computational biology who are constantly nourishing our platform, which learns and improves as we go. However, research and development strategies are set by humans; decisions are made by our experts. Ultimately, approvals of novel drug candidates will be decided by expert regulators based on factual clinical data regardless of whether a given drug had initially been conceived with the help of AI or without.
Hopkins: As mentioned earlier, this shift is already beginning to happen. The economics of AI-driven drug development appears convincing. Most people involved in drug R&D know the industry can (and must) do better than a 96% failure rate. We know we can improve upon the current inefficient process—one in which companies typically screen large numbers of molecules, whittle them down to one or two development candidates, only to watch most of them fail in large-scale human trials years later. Exscientia doesn’t screen any number of molecules. Instead, through our AI-led platform, we design them against pre-determined parameters.
For those who may still hesitate to embrace AI, I would say, look at the numbers. Exscientia’s generative AI-driven drug design process averages about one year versus the industry standard of four to five, during which we typically synthesize around 250 compounds (compared to the 2,500 to 5,000 industry average). There are many serious unmet health needs that are still to be tackled. Knowing there’s a better, more efficient way to solve them, the question I would ask isn’t, can we afford to use AI, but can we afford not to?
Hopkins: Jointly with a small group of like-minded scientists, I founded Exscientia in 2012 with the idea to use generative AI, encoding, and automation to sift through vast datasets and design novel drug candidates faster, in better quality, and at more reasonable investments. I describe us as an AI-driven precision medicine company committed to accelerating the development of the best possible medicines in the fastest and most effective manner. We’re proud of the role we’ve played in pioneering the use of generative AI in our industry, which is evolving in new and exciting ways.
Based on our decade-long experience utilizing integrated AI and automation in drug discovery, we entered the first AI-designed compound into clinical development. To date, we’ve already brought six AI-designed molecules to the clinical stage, with two further currently going through IND-enabling studies. . . . We believe all drugs will be discovered and developed using AI by 2030, and we’re excited to help make it happen.
Meg Rivers is the former managing editor of Pharmaceutical Executive®.