Digital: Drug Discovery

For decades, bringing a drug from concept to patient access has been an expensive, convoluted and often uncertain journey. The traditional drug discovery process relies heavily on trial and error, biological hypotheses and rigid scientific processes. Various challenges significantly prolong the drug discovery process, ultimately slowing the pace of innovation and delivery of treatment for patients. Today, the integration of artificial intelligence into the drug discovery process is accelerating this innovation and fundamentally helping to deliver higher quality treatment to patients at a faster pace than ever before

Tarek Samad at Lundbeck

Before artificial intelligence (AI): the long, costly road to traditional drug discovery

The low probability of success in the traditional drug discovery process is the result of numerous barriers. From identifying the correct biological target through detailed preclinical testing, to navigating the complexities of clinical trials and finally meeting strict regulatory standards, drug discovery has always been a high-stakes process. What’s more, biology can often be unpredictable – usually, nature proves us wrong because what appears promising in theory or in preclinical tests may fail in humans. The reality is humbling; the probability of a compound eventually becoming a successful drug remains discouragingly low. This hasn’t changed much over the past 30 years, despite various technological advancements.

Additionally, the time investment in drug discovery can often be unappealing – accounting for the discovery and concept stages, preclinical and clinical testing, regulatory documentation, and finally, patient access, the whole process usually takes well over a decade. This is particularly critical when developing treatments for diseases with urgent unmet medical needs, such as those affecting the brain. Neuroscience, for example, is particularly complex and time-consuming due to the intricacies of the human brain and the longer timelines required to see meaningful change in conditions like Alzheimer’s disease or depression. Finally, there is the enormous financial burden of drug discovery and development. Depending on the indication and length of trials, developing a single medicine routinely costs between $1-2bn.1

Therefore, given the long timelines and high failure rates, every misstep carries a significant financial cost – making the right choice early on, from identifying the best compound, to targeting the correct disease mechanism, is critical.

Despite these challenges, the end goal of bringing life- changing therapies to patients has always been a worthwhile and rewarding pursuit. Still, the inefficiencies of traditional methods are evidence of a clear need for a new, refreshed approach, and AI is increasingly demonstrating that it can revolutionise this space.

Enter AI: speed, efficiency and better choices

The capabilities of AI mean that all three of these challenges can be addressed to streamline the drug discovery process and ensure safe delivery of treatment to patients, from early idea generation through preclinical and clinical development to regulatory documentation and even patient engagement.

In traditional processes, having multiple hypotheses and testing them can be time-consuming, as it is usually a sequential process and costly. In early discovery, finding the starting-point for new drug candidates has traditionally been based on in-house libraries of potential drug candidates. AI has brought the possibility of in silico screening of ultra-large libraries that can uncover new chemical space. When a starting point has been identified, AI can also identify patterns and generate predictions by running numerous simulations in parallel, suggesting the most promising compounds and helping to deprioritise less promising candidates, long before they reach the lab. For example, if researchers are considering 20 different potential drug candidates, AI can predict which two are most likely to succeed and generative AI (genAI) can enable the design of new candidates with innovative ideas based on the available data. By only selecting the ‘most likely to succeed’ candidates, it is possible to reduce the number of in vitro and in vivo preclinical tests, enhance speed and increase the chances of successfully identifying the lead candidate to take into the clinic.

Turning to clinical development, AI can accelerate patient recruitment, design more efficient clinical trials and streamline regulatory documentation. For example, genAI models can be trained to automatically draft or review regulatory submissions, reducing the so-called ‘white spaces’ that arise while waiting for documentation and approvals, and cutting down on months of delay. This is still an emerging area and such tools are naturally under regulatory scrutiny, with a human in the loop for validation. In clinical trials, AI can identify suitable patient candidates through electronic health records and other digital data sets, all whilst remaining compliant with ethical standards and retaining patient confidentiality. Furthermore, it can help predict disease progression and drug response, enabling smarter trial designs. This ultimately supports putting patients first throughout the entire clinical trial process.

By streamlining the traditional drug discovery process through reduced failures and improved accuracy and efficiency, AI can significantly lower the overall cost of drug development from preclinical design all the way to commercial launch, because it reduces the need for extensive trial-and-error in the lab and accelerates the decision-making process. Importantly, this cost efficiency does not mean cutting corners – rather, it enables pharmaceutical companies to allocate resources more effectively and reinvest cost savings into additional innovation. AI is also starting to show potential in understanding patient needs and reaching them more efficiently through digital engagement platforms. While fully AI-powered clinical trials are still a long way away, without a doubt, AI is becoming a powerful resource for almost every phase of drug development.

The power of partnership: why collaboration is key

Given the current pace of AI innovation, investing in internal advanced AI platforms can be a risky and costly move for pharma companies, as they can take years to develop and then quickly become outdated. As such, many companies are prioritising partnerships with best-in-class external providers to supplement internal efforts, which offers several advantages.

By working with external partners, companies are able to quickly pivot and explore new opportunities without allocating large amounts of capital to rapidly ageing platforms and remaining locked in to any single technology.

Investing heavily in building broad AI solutions risks becoming a case of searching for problems to solve in order to justify the investment. Instead, by addressing a clearly defined challenge, and then seeking the most appropriate AI partner

to address it, solutions can remain practical, efficient and directly relevant. Outsourcing solutions in this manner allows the expert partners to handle the tech-heavy lifting, and means that resources can be scaled up or down as needed.

Neuroscience meets AI: now and beyond

AI models rely heavily on high quality data sets to make accurate predictions and work effectively.

Therefore, it is extremely important to invest in the necessary data infrastructure to ensure that we are well-equipped to supply AI models with the correct information and data to work effectively.

Looking ahead, as prediction models improve, scientists will be able to predict not only how a molecule will interact with its target, but also how it will behave in the human body, including potential safety risks and efficacy across diverse populations. As technologies and data-sharing ecosystems have evolved, the question of using AI in healthcare has shifted from ‘if’ to ‘where’ and ‘when’. Keep in mind that biological systems are complex, and AI models may not always fully understand and predict the interactions within these systems. As AI models are trained on specific data sets, they may not generalise well to real-world scenarios. This can lead to oversimplified models that do not capture the intricacies or full extent of human biology. Therefore, real-world verification remains paramount in drug discovery, and the need for empirical testing to confirm the value of AI-suggested targets, drugs or approaches will continue to be relied upon. Nevertheless, AI is allowing researchers to pursue therapeutic areas that have traditionally been considered too complex or high risk, increasing the potential for identifying novel targets and developing treatment strategies to address any so far unmet patient needs, because the quality of candidates that are advanced into clinical trials will be improved, maximising patient benefit by speeding up access to treatment.

In neuroscience specifically, AI can speed up the process of designing and optimising a molecule that targets one or multiple receptors in the brain – a vital functionality in developing innovative treatments for neurological diseases – which can then be assessed by internal researchers to determine the most impactful molecule in a human context.

The pharmaceutical industry is certainly at an inflection point. For decades, drug discovery has been slow, costly and uncertain. But with the advancement of AI, and with the strategic adoption of AI into pharmaceutical companies, the industry is going through momentous change. Maintaining a focused, question-first approach to AI partnerships to equip experts with better tools to enhance drug discovery, rather than replace or automate the process, is revolutionising the way treatment is delivered to patients, improving not only the quality of treatment but also the outcome for them. As the technology continues to evolve, the industry must focus on leading not through scale but through precision, flexibility and a clear understanding of where AI can make meaningful impact.

Looking further ahead, the emergence of agentic AI – systems capable of autonomously setting and pursuing scientific goals – could further transform drug discovery. These systems may one day assist researchers not just by analysing data, but by independently proposing novel hypotheses, designing experiments and adapting strategies in real time based on outcomes. While still in early stages, agentic AI holds the potential to act as a true collaborator in science, accelerating innovation in ways that are currently difficult to imagine. As with all powerful technologies, this will require careful oversight, ethical frameworks, and robust validation to ensure safety and reliability.

The future of drug discovery won’t be built by AI alone, but by smarter science, better questions, better answers and, ultimately, better outcomes for those who need them most.

Reference
1. Visit: cbo.gov/publication/57126