05 August 2025

Aligning AI Innovation with IP Strategy in Drug Discovery

Investment Surges in AI-Driven Drug Discovery

Artificial intelligence (AI) has emerged as one of the most transformative technologies in recent years. The European Medtech sector saw a surge in investor interest in early 2025, dominated by AI-powered solutions. In Q1 alone, AI startups secured 25% of all European venture capital funding, with AI for drug discovery emerging as one of the leading segments.

Earlier this year, the UK government announced a £82.6 million investment into cancer research using AI, showing a dedication to harnessing the power of AI for cancer care and drug discovery. In June, the Nuffield Department of Medicine announced a new consortium, based in Oxfordshire, which will generate the world’s largest trove of data on how drugs interact with proteins for training AI models. 20 times larger than anything collected over the last 50 years, this collection will allegedly cut drug discovery costs by up to £100 billion.

Confidence in the potential of AI to solve healthcare’s greatest problems is growing, especially in areas like small-molecule drug and antibody design. With pressure to deliver faster, more targeted therapies, AI is becoming a central engine of biomedical innovation.

AI’s Transformative Role in Drug Discovery

AI is revolutionising how we discover new drugs by unlocking speed, scalability and novel insight. Traditionally, identifying new drug targets relies on a mix of intuition, laborious experimentation and trial-and-error. Pharmaceutical companies typically take 10 to 15 years to bring a single drug to market, which can cost up to $2 billion. Despite this effort, only about 10% of candidates entering the trial pipeline eventually succeed.

These are concerning statistics, but AI’s ability to sift through vast biological datasets and carry out predictive modelling could be the answer. AI can assist at virtually every stage of the small molecule drug discovery pipeline, including target identification and validation, hit discovery, lead optimisation, and preclinical assessment.

Reshaping structural biology with AI

The use of AI in structural biology has become crucial for modern drug discovery. AlphaFold, developed by DeepMind, represents one of the most groundbreaking achievements of AI. Its creators, Demis Hassabis and John Jumper, were awarded one half of the 2024 Nobel Prize in Chemistry “for protein structure prediction” in recognition of their work on AlphaFold.

Traditionally, tertiary protein structures have been determined through complex and time-consuming techniques such as X-ray crystallography. In contrast, AlphaFold enables the prediction of protein structures based on amino acid sequences, which are readily available in different databases. AlphaFold has been widely adopted by the scientific community and has become an indispensable tool in structural biology since its public release. It enables medicinal chemists and structural biologists to identify binding pockets, model ligand interactions, and perform in silico docking studies, even for proteins previously considered “undruggable” due to lack of structural data.

The recently released AlphaFold 3 model further advances the field by improving the prediction of protein-ligand interactions, including the binding of antibodies to target proteins. This enhanced capability is expected to significantly accelerate the design and optimisation of therapeutic antibodies, which now represent a critical class of biologic drugs.

AI for target identification and drug design

AI plays a pivotal role in both target identification and drug design. Enabling researchers to identify novel or previously overlooked drug targets, AI algorithms can mine genomic, transcriptomic and proteomic data to prioritise genes or proteins implicated in disease pathways. AI also accelerates the discovery of lead compounds by predicting molecular properties and optimising chemical structures.

One of the most promising recent examples of an AI-discovered drug is rentosertib, which is a small-molecule inhibitor of Traf2- and Nck-interacting kinase (TNIK). TNIK was identified as a potential therapeutic target for idiopathic pulmonary fibrosis (IPF) through AI-powered analysis of gene expression datasets profiling the tissue of patients with IPF. A separate AI platform then designed and optimised the small molecule drug. Remarkably, it took less than 30 months to progress from target discovery to the completion of Phase I clinical trials. In a recently conducted Phase 2a trial, preliminary results showed that rentosertib was well tolerated and led to significant improvements in patients’ conditions compared to the placebo group.

Unlocking the hidden potential of existing drugs through AI

AI can also help identify potential medical uses for existing drugs, and this approach can save time by reducing the need to optimise drug structures and address potential safety issues.

Baricitinib is a Janus kinase (JAK1/2) inhibitor originally developed for the treatment of rheumatoid arthritis. During the COVID-19 outbreak, researchers used an AI-driven knowledge graph platform to explore existing drugs that could potentially be repurposed to combat SARS-CoV-2. Baricitinib was found to have both antiviral and anti-inflammatory properties which could be useful in treating COVID-19, and the FDA approved it for use in patients with severe COVID-19 soon after.

Intellectual Property: The Strategic Imperative

In the rapidly evolving field of AI-driven drug discovery, intellectual property is more important than ever to attract investment and protect your ideas. Securing patents not only protects novel molecules and AI platforms but also increases a company’s value and positions them as a leader in the competitive marketplace of drug discovery.

A 2024 patent landscape report recorded 1,087 global filings related to AI-enabled small-molecule discovery between 2002 and 2024. Recently, patent filings as well as pending applications in this area have surged, reflecting exciting technological progress.

However, unlike patenting new drugs in the traditional pharmaceutical industry, which could be relatively straight-forward, bringing AI into the mix could complicate things.

AI as the inventor

Despite the growing role of AI in research and development, both the European Patent Office (EPO) and UK Intellectual Property Office (UKIPO) clearly stipulate that AI cannot be named as an inventor on a patent application. In the landmark DABUS cases, where Dr. Stephen Thaler attempted to name AI system, DABUS, as the sole inventor, both jurisdictions rejected the applications on the grounds that only a natural person can be legally named as an inventor. Although AI-generated inventions may still be patentable if a human (such as the deviser of an AI model) claims inventorship, AI itself cannot hold legal rights or be recognised as the originator of a patentable invention.

AI-generated inventions

Inventions such as drug molecules and antibodies designed and/or optimised by AI are patentable under UK and EPO law, provided that they meet the standard legal criteria of novelty, inventive step and industrial applicability. Furthermore, experimental data showing that the AI-designed drugs can achieve specific “technical effects”, such as enhanced efficacy, specificity or binding affinity, is key to securing the grant of a patent.

As with any invention, companies should file early to secure protection, as well as conduct thorough IP searches to avoid infringing on others’ rights. Monitoring existing patent filings in biomarkers and therapeutic targets also allows companies to focus their R&D efforts on drug candidates that can satisfy the patentability requirements of novelty and inventive step.

AI as the invention

In addition, it is advisable to consider obtaining patent protection for any novel AI system which identifies biological targets or designs molecules, not just the output. 

Along the same line as algorithms and software, intellectual property law in the UK and Europe views AI models per se as of an abstract mathematical nature and therefore not patentable. However, an AI or machine learning invention may be patentable if it produces a technical effect that serves a technical purpose, either by its application to a field of technology or by being adapted to a specific technical implementation. 

In this regard, under EPO practice, AI inventions applied to the specific field of drug discovery may be patentable, especially if such AI models solve clearly defined drug development problems (e.g. improving binding affinity or reducing toxicity).  Patent protection may also be available for so-called core AI inventions relying on developments of the fundamental underlying AI techniques, rather than application of an AI model to a particular technical field such as drug discovery.  The UK follows a broadly similar approach, with courts applying the Aerotel/Macrossan test, originally devised in the context of general computer-implemented inventions, to assess whether an AI invention is a patentable technical contribution.

Thus, while AI cannot be named as an inventor on a patent application, it can certainly be the subject of patent protection if it contributes to the technical character of an invention.  Care must, however, be taken in drafting any patent application directed to such AI inventions.

Conclusion

As AI transforms how we discover and design new medicines, intellectual property becomes a critical pillar for translating technological breakthroughs into lasting competitive advantage.

Given the fast-moving nature of the field, a robust, multi-faceted IP approach is crucial. Other things to consider are leveraging trade secrets as well as patent protection and engaging in strategic licensing, open innovation and partnerships to help ease costs, allow broader access to vital data for AI training, and accelerate development. Most importantly, integrating a comprehensive IP strategy into your research and business activities from the outset will position your company to thrive in this dynamic landscape.