How Is AI Being Used in Drug Discovery?

The field of drug discovery faces many time-honored challenges. These include high costs, long development times and high rates of failure. The rise of artificial intelligence (AI) technology has been hailed as a potentially transformative tool to tackle these drug discovery challenges; but is it living up to the hype?

To explore further, Technology Networks spoke with a handful of drug discovery experts who apply AI to a range of drug discovery stages to learn more about its impact in the pharmaceutical space.

AI throughout the drug development pipeline

In its simplest form, AI is “the utilization of cognitive technologies such as algorithms, machine learning and robotic process automation,” according to Yuan Wang, head of research analytics at the Belgian pharmaceutical company UCB.

AI boasts the ability to mimic human cognition, decision-making and action-taking capabilities – something that drug discovery scientists can put to good use. With the recent explosion of interest in AI, and the growth of its power, excitement among drug developers began to build over its potential in the pharmaceutical space. AI technology could sidestep the many pitfalls and hurdles inherent to drug development, speeding up the process and ultimately helping to make safer and more effective drugs.

“In traditional drug discovery, it takes over 10 years and costs around $2 billion to bring one drug to market – and 90% of drug candidates fail during human trials,” explained Dr. Alex Zhavoronkov, founder and CEO of the AI-driven biotech company Insilico Medicine. “This high cost and slow speed are preventing new life-saving medications from reaching patients.”

AI can come in handy at many different points in the drug discovery process, from initial target identification to optimizing clinical trials.

“AI can help us to find patterns in data and, increasingly, to infer causality,” said Roger Palframan, former head of US research at UCB  and now chief scientific officer at Stealth Biotech. “This can help us prioritize from a long list of potential drug target options and help us make better choices about the targets to go for.”

By deepening our understanding of disease biology in this way, AI can help by identifying suitable drug targets and informing how the drug compounds are designed.

“In the pharmaceutical field, AI has opened remarkable possibilities from understanding disease pathology to better-designed clinical trials,” Wang explained. “The utilization of AI can bring undeniable value to patients as it facilitates efficient analysis of genetic data, disease pathways and gene sequences.”

The overarching goal is to make a real impact on human health and how we treat disease by reducing the time it takes to get medicines from initial development through to patients.

“AI could potentially help us to find molecules that we would not have found using, say, ‘brute force’ high-throughput screening approaches,” said Palframan. “So, we're using AI today for small molecule discovery to help us make compounds faster, to choose more quickly which compounds to make and to reduce the number of compounds we're making.”

“Harnessing the power of AI in our drug discovery efforts has exciting potential for improving patient outcomes and transforming the landscape of severe disease treatment by speeding drug discovery and developing a more personalized approach,” said Wang.

Enhancing clinical trials

But AI’s applications don’t stop at compound discovery; it can also be applied to great effect further down the drug development pipeline.

After a lead compound is developed, the next steps are pre-clinical trials in animals and then clinical trials in humans. AI can enhance the design and performance of clinical trials by optimizing treatment schedules, participant recruitment and data accessibility.

For example, though some researchers remark that many patients show a keen interest in taking part in clinical trials, currently just 5% of eligible patients participate in clinical research.

There are a few different ways that AI can be helpful in this scenario, explained Miles Witham, professor of clinical trials for older people at Newcastle University: “One is helping us to find and recruit the right patients. We know patients want to take part in clinical trials, and sometimes it's quite hard to match patients and trials together."

AI tools can help speed up the recruitment of eligible patients and alert both healthcare professionals and patients to new clinical trial opportunities. They could also save huge amounts of time, resources and volunteers needed for new trials by improving not just their design, but also through making better use of the data we already have from existing trials.

“Utilizing AI capabilities for analyzing huge volumes of data and machine learning algorithms, we can better utilize the information that we have from previous clinical trials,” said Wang. “This learning can inform future trials and identify promising solutions for the patients we serve.”

“By leveraging AI's capabilities that can identify eligible participants, optimize trial design and utilize real-world data, researchers can overcome challenges associated with patient recruitment, enhance diversity and inclusivity and overall help improve a clinical trial,” Wang continued. “The integration of AI technology in clinical trials holds immense potential for advancing medical research and delivering better healthcare outcomes for patients worldwide.”

This 5% figure of total eligible participants taking part in trials is even lower in underrepresented groups, such as ethnic minorities or the elderly. This is also where the role of AI can come into play, as Wang explained: “AI can help address this issue by identifying gaps in representation and suggesting strategies to ensure diverse participation. By including a broader range of individuals in clinical trials, researchers can gain insights into the drug’s effectiveness and safety in different populations.”

Impacting personalized and precision medicine

AI is also having an impact on personalized medicine – in other words, the tailoring of treatments to individual patients based on their genetic makeup. This shift to a more precise approach for drug design could help lead to more personalized treatments, which come with the added benefits of helping therapies to be less toxic and more effective.

AI can touch on personalized and precision medicine in a few different ways, such as through analyzing huge datasets to identify trends and patterns according to a patient’s genome, or by optimizing the combination of drugs that a cancer patient may take.

“AI also has the potential to assist in optimizing combination therapies, which involve using multiple drugs together,” Wang explained. “By analyzing diverse datasets and patient-specific characteristics, AI algorithms can predict the synergistic effects of different drug combinations and identify optimal dosage regimens. This can lead to the development of more effective treatment strategies, particularly in complex diseases like cancer.”

Furthermore, the benefits can go beyond tailoring medicines; it could also identify groups of patients more likely to respond well to a particular therapy.

“AI can help us to define the patient population we're looking for across many diseases,” said Palframan. “This is why when we talk about ‘the right target, the right drug and the right patient’, the right target and the right patient go together. If we can find a more homogenous, molecularly better-defined patient population, that would help us find better causality concerning the target. This is important, as failure of mechanism is the number one reason for failures of therapeutics.”

Does it live up to its promise?

AI became “mainstream” with the rise in popularity of ChatGPT and the subsequent explosion of AI tools. For drug discovery, it was largely hoped that AI would revolutionize and overhaul the process, with scientists even using AI tools to write articles on the promise of AI in drug discovery. But has it lived up to the hype, or is its success being barred by resistance to change?

“Perhaps the major limitation is remaining skepticism in the industry, although AI is increasingly being explored by most big pharmaceutical companies eager to advance their own programs and accelerate the pace of drug discovery,” said Zhavoronkov. “There is still no AI-generated small molecule drug treating patients, which many are waiting for.”

So, AI in its current form may not the Holy Grail – it is still dependent on high-quality data to produce meaningful results.

“Any value in AI’s ability to synthesize huge amounts of information and transform it into actionable insights and lifesaving medicine will be entirely dependent on the quality of the data being fed in,” said Luke Cox, CEO at biotech company Impulsonics. “Otherwise, we will simply be feeding these models noise and using a lot of energy to generate not a lot of useful information. Or to put it another way; garbage in, garbage out.”

Many ask, to what extent could AI replace real, human scientists? There is increasing concern that AI could be cutting the number of available jobs in some industries. At present, though it seems AI can provide significant benefits to the drug discovery pipeline, it is unlikely to make scientists obsolete just yet, augmenting scientists’ work but still requiring human oversight. In other words, AI can provide insights and predictions, but it remains up to the scientists to interpret and validate them.

“Biological systems are highly complex, and our understanding of their intricacies is still evolving. AI models in drug design need to account for this complexity, considering the interactions between multiple targets, pathways and physiological responses,” said Wang. “For this reason, AI should be used to inform researchers during the drug discovery process rather than replace them.”

What could the future hold?

The future of AI in drug discovery offers promise and potential across the pharmaceutical landscape to benefit patients.

“With AI’s capabilities evolving at an unprecedented rate there is no telling where the future of AI in the drug discovery space could lead,” said Wang.

“As well as shortening the time from the lab to the patient, improving clinical trial diversity and moving medicine in a more personalized direction, we hope to see AI informing drug safety and toxicity predictions. This can enhance patient safety and increase the efficiency of the drug development process.”

As scientists continue to innovate, AI will likely become embedded into research and development workflows as it increasingly becomes a larger part of our daily lives, Palframan explained: “I think that's how AI will be used in drug discovery; it will become part of what we do and very normal. That will come through seamless integration, especially between the wet lab and the dry lab in silico; that's where the real value will be created.”

“Increasingly, I believe it will be important to provide pharmaceutical companies with full end-to-end capability, from target identification to drug design to clinical trial prediction in order to create one seamless pipeline. I also believe robotics will increasingly work in concert with AI to bring additional speed and capacity to drug development,” said Zhavoronkov.

“I think having a vision for the future is great, but I'm not entirely sure how we get there. Therefore, as this evolves quickly, how do we set ourselves up for success to be agile and not constrain the power of AI, especially in early research? How do we free the AI to be able to impact discovery? How do we apply the appropriate risk and framework to some areas, such as patient-facing AI where it naturally needs to be more regulated? But also, how do we free it in other areas to be able to maximize its potential?” said Palframan.

While the future of AI in drug discovery is bright, it is clear that realizing its full potential will require both further innovation and careful planning.