Just as structure-based drug design, combinatorial chemistry, Cryo-EM, and AlphaFold significantly advanced their respective eras, artificial intelligence (AI) now electrifies medicinal chemistry. While the fusion of AI and machine learning (ML) undeniably holds the power to accelerate drug target identification and new medicine discovery, their most significant opportunity lies in data interpretation.

This article explores the specific areas where AI and ML provide a genuine advantage in drug discovery. We'll also offer our perspective on the challenges ahead and the key aspects to watch.

Before we start, let’s define a few terms:

Artificial intelligence (AI): an umbrella term for a range of advanced computational and modelling techniques that analyse and learn from, often large and complex, data sources, and can generate insights or perform tasks that would typically require human-level intelligence at a scale and speed beyond human capability.
Machine learning (ML): a subfield of AI that focuses on developing algorithms and statistical models enabling computers to learn from and make new predictions or decisions based on data.

Both definitions taken from the Wellcome Trust and Boston Consulting Group’s report “Unlocking the potential of AI in Drug Discovery”.

AI in Drug Discovery: It’s Not Always ChatGPT

It's important to understand that applying AI to drug discovery isn't typically about asking ChatGPT to explore new medicines for breast cancer treatments - despite a few companies venturing into that territory. In many cases, the AI tools and techniques we're discussing are entirely different from platforms like ChatGPT.

AI and Machine Learning in Drug Discovery: Specialized Tools, Expert Use

Pharmaceutical AI and ML software represents a highly specialized area. These programs are built and trained on specific, carefully controlled, and regulated data to deliver clearly defined results. Notable examples include Merck's ADDISONTM, Schrödinger’s Jaguar, and Scripps’ AutoDock, all requiring specialist knowledge for optimal performance.

A review by the Wellcome Trust and Boston Consulting Group identifies five major ways AI is transforming drug discovery: from understanding diseases to designing and optimising small molecules, vaccines, and antibodies, as well as predicting and analysing safety and toxicity. These applications are yielding valuable cost and time efficiencies. However, the substantial computational power required raises environmental concerns, and the field faces hurdles like uneven development, notably a lack of private funding for less lucrative therapeutic areas and underinvestment in AI for middle- and low-income nations.

AI Solutions for Key Drug Development Challenges

Key pain points in drug development, including discovering new ways drugs bind to targets and achieving selectivity to prevent off-target effects, are being effectively addressed by AI. This technology is proving to be exceptionally useful for analyzing intricate biological systems, pinpointing disease biomarkers and drug targets, simulating how drugs interact with their targets, forecasting the safety and effectiveness of potential medicines, and managing clinical trials (Figure 1). A notable example is the discovery of NDI-034858, a selective TYK2 allosteric inhibitor for moderate to severe psoriasis, by Nimbus Therapeutics using Schrödinger’s FEP+ physics-based model. The significance of this AI-driven discovery is underscored by Takeda's USD 6 billion acquisition of NDI-034858 (now TAK-279) in December 2022.

Figure 1: Schematic to show the uses of AI within the drug discovery process. Taken from Zhang, K., Yang, X., Wang, Y. *et al.* Artificial intelligence in drug development. *Nat Med* 31, 45–59 (2025). https://doi.org/10.1038/s41591-024-03434-4

The AI drug discovery landscape includes companies like BenevolentAI that partner with institutions for development. However, risks remain, and some AI-based ventures face underperformance. This is because, despite AI's benefits, the resulting drug candidates can still suffer from issues such as unexpected off-target interactions, toxicity, and poor pharmacokinetics.

AI for Faster and More Efficient Drug Development: Recent Advances

Significant AI breakthroughs in the last five years are transforming drug discovery. AlphaFold, a deep learning algorithm, accurately predicts protein structures with near-experimental accuracy, enabling drug discovery companies to access previously uncharacterized targets. While few AI-discovered drugs are yet in the clinic, early data suggests 25-50% time and cost savings, alongside improved chances of finding a successful and novel therapeutic agent of a target. Additionally, AI is facilitating drug repurposing and candidate repositioning efforts, including in Alzheimer’s disease (AD) treatment.

The rapid discovery and clinical trial entry of Relay Therapeutics’ selective FGFR2 inhibitor, RLY-4008 (lirafugratinib), Figure 2, showcases the power of AI in drug development. Long-timescale molecular dynamics simulations were used to identify the differential motions of FGFR2 and FGFR1 in their P-loops, which are located adjacent to the orthosteric site. This understanding enabled the rational design of inhibitors with high FGFR2 selectivity. The resulting covalent binders demonstrated significant in vitro selectivity, approximately 250-fold over FGFR1 and 5,000-fold over FGFR4.

Figure 2: RLY-4008, lirafugratinib, an FGFR2 inhibitor that exploited molecular dynamics simulations during the initial design phase.

The rapid progression of RLY-4008, an FGFR2 inhibitor, from initial identification to clinical trials in only four years demonstrates remarkable efficiency. This was further recognized in January 2022 when the FDA granted RLY-4008 orphan drug designation for cholangiocarcinoma. Compared to the average 12–15-year drug discovery process (Figure 3), this timeline showcases a substantial improvement.

Figure 3: The potential impact of AI upon drug discovery time and cost. Image taken from Wellcome Trust and Boston Consulting Group’s report “Unlocking the potential of AI in Drug Discovery”.

Making AI Work for Drug Discovery: Essential Success Factors

Several elements are critical for successful AI-driven drug discovery, with clear communication and seamless integration between scientific and technological teams being paramount. This interdisciplinary collaboration is especially vital during target identification and compound generation to ensure thorough sense-checking, as technology-focused team members may lack deep scientific knowledge and vice versa.

Additional crucial factors include:

Robust data access and standardization for storage and future use
The collection or generation of high-quality, relevant data
Rigorous algorithm training and validation, particularly addressing the challenge of obtaining sufficiently large and high-quality datasets (especially critical for biological systems where inter- and intra-laboratory data variability can be significant, such as with microsome batch data)
Careful consideration of ethical implications, notably patient data protection.

In Conclusion: Key Takeaways on AI in Drug Discovery

The integration of AI into drug discovery and design is a dynamic and accelerating field, evidenced by the increasing number of AI-derived compounds reaching the clinic. Despite its advancements, AI is not a complete solution, and human involvement is still required. This includes essential laboratory validation of clinical candidates (often via CROs) and the generation of the necessary data to train effective AI algorithms.

Furthermore, the structure of medicinal chemistry and drug discovery teams is becoming increasingly interdisciplinary. The traditional separation of medicinal chemists, biologists, and computational chemists is becoming less common, leading to more integrated teams working collaboratively, often co-located, with a shared objective but diverse knowledge bases. Therefore, clear and concise communication is more important than ever, regardless of the audience – from internal teams to external collaborators, prospective clients, or the wider public.

Contact ConsultaChem today to learn how our communication solutions can benefit your organization.

AI in Drug Discovery: More than just ChatGPT