Smarter tools, smarter solutions – how AI is transforming drug discovery

AI is already playing a fundamental role in every stage of the drug discovery process – from ideation all the way through to drugs reaching patients.

This integration is accelerating innovation and enabling the delivery of higher-quality medicines at a faster rate.

We have become accustomed to drug discovery and development being a slow, costly process fraught with uncertainty. Estimates have put the cost of taking a drug from concept to market at between $1-2 billion, with the process taking more than a decade.

What’s more, despite the numerous technological and scientific advancements we have witnessed over recent decades, the probability of a compound actually making it to market is extremely low.

Taking all these factors into account, it is imperative that we continue to find new ways to de-risk the drug development process.

Through the considered use of AI, our experience in Lundbeck is that we can eliminate a number of the inefficiencies in traditional drug development methods to make sure we make smarter choices earlier on.

The benefits of using AI can be seen right from the early discovery phase, where the traditional method of testing multiple hypotheses in a sequential process is often both time-consuming and costly.

Through the use of AI models, multiple simulations can be run simultaneously, identifying patterns and generating predictions that can determine which compounds are most promising, as well as those that should be deprioritised.

The ability to conduct this process before entering the lab is hugely beneficial, as it reduces the number of in vitro and in vivo tests that need to be conducted and allows compounds to be de-risked at an earlier stage.

Once a compound enters clinical development, AI can streamline a range of processes. On the documentation side, AI models are being trained to draft or review regulatory submissions to reduce the often lengthy process of waiting for approvals.

Given this is still an emerging application for AI, tools like these are naturally subject to regulatory scrutiny, and a human remains involved in the process to validate the model’s outputs.

In clinical trials, AI can address common pain points in the process such as patient recruitment and retention. Using electronic health records and other data sets, AI can identify suitable patients to enrol in a trial, without compromising on patient confidentiality, data protection regulations or ethical standards.

From a patient perspective, we are starting to see some promising uses of AI to help us gain a better understanding of patient needs, and to reach patients more efficiently through various digital engagement platforms.

Naturally, patients will still value the opportunity for human interaction, so rather than replacing humans with AI tools, it is important that we take the time to consider how they can complement one another to ensure patients have a smooth and efficient trial experience.

As the applications of AI evolve, agentic AI, whereby systems are able to autonomously set and pursue scientific goals, has the potential to further transform the way in which we approach drug discovery.

This approach would see AI act as more of a collaborator for researchers, rather than a tool, by moving away from simply analysing data to suggesting hypotheses, designing trials and adapting protocols based on real-time data and outcomes.

Naturally, with new and exciting solutions constantly emerging, companies want to get on board with using AI as quickly as possible. However, as with any technology, investing in and developing in-house platforms can be a time-consuming, risky and costly process.

What’s more, in such a fast-moving field, any newly developed platform can quickly become outdated and obsolete.

As such, it often makes sense for pharma companies to partner with best-in-class external partners. We have good experience with that ourselves, both in California with Iambic Therapeutics and in Denmark with DCAI.

This provides the ability to explore different opportunities and capabilities without becoming locked in to a single technology or spending large amounts of capital on platforms that are not fit for purpose or ageing.

Furthermore, using external providers means that the use of certain resources or tools can be dialled up or down as needed. This avoids the common problem experienced when companies invest heavily in developing broad AI solutions without clearly defining the challenge they are trying to address, and therefore end up trying to find problems to solve in order to justify this investment.

Adopting this question-first approach to choosing AI tools or partners, rather than simply trying to replace or automate existing processes, will allow companies to find the right tool for them.

One area that it is hugely important to invest in from an AI perspective is high-quality data infrastructure.

The output of an AI model can only be as accurate and effective as the data that is inputted into it, and therefore it is vital to ensure that we are able to supply these models with the right data.

It is also important to remember that AI may not always be able to fully understand the intricacies of biological systems and the interactions that take place within them.

Furthermore, AI models are trained on specific data sets, which can lead to oversimplified models that are not able to generalise or reflect real-world scenarios.  This means that targets, drugs or approaches suggested by an AI model still need to be empirically tested in real-world scenarios to validate the hypotheses generated.

With high-quality data, AI is also enabling researchers to focus on targeting therapeutic areas that have historically been considered to be too complex or high risk to have any chance of success.

With these new capabilities, we should be able to address existing areas of unmet need by identifying novel targets or treatment strategies that would not have seemed feasible in the past.

By accelerating the decision-making process and reducing the likelihood of trial failures, AI can substantially lower the cost of drug development while improving the accuracy and efficiency of the traditional drug discovery process.

However, it is not just AI that will transform the future of drug discovery, but smarter science, which means asking the right questions to enable these tools to find the solutions to current pain points in the process.