In a milestone for science and technology, the 2024 Nobel Prize in Chemistry was awarded to the DeepMind team for their work on AlphaFold, an artificial intelligence (AI) system that accurately predicts protein structures. By solving one of biology’s most fundamental problems, AlphaFold demonstrated the potential of AI to advance drug discovery.
At the same time, advances in generative AI and protein modelling are pushing the boundaries of what is possible in drug design, while virtual cell simulations hint at a future where some experiments could be conducted in an entirely digital manner. Yet, for all its potential, the use of AI in drug discovery is not without its challenges.
Go deeper with GlobalData
Data scarcity, biological complexity, and regulatory concerns still present significant hurdles to overcome. And while the hype surrounding AI garners steam, it often overshadows the slower, incremental progress required to bring real-world solutions to market.
According to a survey by GlobalData, the parent company of Pharmaceutical Technology, AI is considered the most disruptive technology among businesses, including in the healthcare industry. As we look towards 2025 and beyond, the question arises: how close are we to seeing AI’s potential in delivering faster, more affordable, and more precise therapies?
Where AI is making an impact
Sara Choi, biotech investor and partner at the Palo Alto, California-based Wing Venture Capital is very optimistic. “I think there’s going to be three times the number of approved drugs in the next ten years, all thanks to these innovations that are happening in the early R&D process,” she says.
Choi says AI adoption is already broad and impactful to the drug development process. “Across the board, they [C suite executives at major pharma companies] say they are already using AI up and down the entire development stack,” she says. This includes early discovery and target identification uses—areas where AI is particularly strong at analysing large datasets to uncover new biological targets. Beyond that, AI is also helping with protein structure prediction, optimising molecular interactions, and even scaling up production to ensure compounds are commercially viable, said Choi.
How well do you really know your competitors?
Access the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain competitive edge.
Thank you!
Your download email will arrive shortly
Not ready to buy yet? Download a free sample
We are confident about the unique quality of our Company Profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the below form
By GlobalDataAI’s role extends into clinical trials as well, with it already being used to enhance trial design, feasibility and site selection, as well as patient recruitment and retention, data analysis and regulatory submission and review. A clinical trial patient matching tool called TrialSearch AI, which operates on a Large Language Model (LLM), was able to reduce pre-screening checking time for physicians by 90%.
Despite this progress, there are limits to what AI can currently achieve. For example, while AI excels at sifting through large volumes of data to identify potential drug candidates, the actual process of moving from discovery to market still relies heavily on human expertise and experimental validation.
Finding the right data is a challenge
One of the biggest challenges in fully incorporating AI in drug discovery is the need for large-scale, high-quality datasets to train AI models. Unlike fields where data is abundant, biological data is often expensive and time-consuming to generate.
Choi says data is the foundation for machine learning. But in biotech, data generation happens through experiments, which are slow and resource intensive. Existing datasets are often small and don’t capture the full complexity of biology.
Adityo Prakash, CEO of Fremont, California-based pharma company Verseon says that AI trained on the existing data can deliver small tweaks and new uses for previously synthesised drug molecules. “But, AI cannot explore the decillion possibilities for which no experimentally derived training data exists. Yet, it is among those decillion possibilities that we will find the breakthrough treatments of the future.”
Prakash’s company is trying to get around AI’s “data problem” for drug discovery using molecular physics. First, novel drug candidates can be designed “using the rules of physics and chemistry” and their binding affinities for disease-associated target proteins can be identified without the need for prior training data. “After we make these completely novel molecules in the lab and generate experimental data, AI can train on this new dataset to find variants as part of the candidate optimisation process and help us identify the best ones to advance to clinical trials,” outlines Prakash.
Emerging directions in AI research
Beyond established applications, new AI-driven approaches are gaining attention in the pharmaceutical industry. One such area is the development of “virtual cells,” which aim to digitally model biological systems in high detail. Demis Hassabis, DeepMind CEO and AlphaFold co-creator, describes this as a long-term goal. “My dream is to eventually have virtual cells, like a simulation of a virtual cell. We’re maybe ten years away from that,” he said at the Financial Times (FT) Pharma and Biotech Summit in London on 6 November 2024.
Virtual cells could allow researchers to run in silico experiments, simulating the effects of drugs or genetic modifications without needing to conduct costly and time-intensive laboratory tests. Regina Barzilay PhD, AI faculty lead at Massachusetts Institute of Technology (MIT), sees this as a way to address a fundamental challenge in biology – the complexity of translating experimental results across species.
Many failures in drug development stem from differences between experimental models and humans, explains Barzilay. Virtual models could provide insights into safety and efficacy of a drug before moving to clinical trials, potentially reducing costs and failure rates.
Generative AI is another emerging field with significant implications. By designing novel proteins or small molecules tailored to specific therapeutic goals, generative models are helping researchers create targeted solutions for complex diseases. Barzilay emphasises that while these tools are still evolving, they are already improving the ability to predict molecular interactions and design effective compounds.
However, she also points out a major limitation: understanding disease signatures. “In order to know how to drug something, you need to understand what’s wrong at the molecular level … this is an area that is still an active area of research.”
Cost reduction drives investment priorities
One of the most promising areas for AI – and one that has seen a lot of investment according to Choi – is its ability to narrow the search space for drug candidates. By helping researchers focus on the most viable options early in the process, AI can reduce costs and accelerate timelines.
While AI is making early discovery more efficient, Choi stresses that the industry is still far from a fully AI-driven process, where “we can just use a machine”.
AI’s greatest potential lies in its ability to make drug discovery more efficient and cost-effective. By reducing uncertainty and improving decision-making early in the process, AI can lower development costs and increase success rates.
This acceleration could have a measurable impact on drug development economics. Choi estimates that the cost of reaching a Phase I readout could drop from over $100m, in some cases, to $70m. Lower costs would enable more companies to pursue innovative therapies, potentially increasing the number of new drugs approved in the coming years.
