The future of cell therapy: AI has a role to play in breaking today’s barriers

Rafael Carazo Salas, CEO and Founder of CellVoyant, explores AI’s potential for cell therapies. AI is already making big strides in biotechnology and healthcare, from its remarkable accuracy in detecting breast cancer from mammograms, to AI platforms such as DeepMind’s AlphaFold revolutionising protein structure prediction. One of the most promising areas for AI is cell therapy – where it can help to speed up research, optimise manufacturing, and significantly lower the cost of finding and developing a new therapy. Despite its vast potential to help treat and potentially cure currently untreatable conditions, cell therapy faces significant hurdles which have to date stopped them from being used widely. One major challenge is producing enough cells for large-scale production, while maintaining quality and consistency. Other key challenges include the time needed to test new cell therapy candidates, as well as regulation being stringent and time consuming. The need for specialised facilities and advanced equipment adds a further layer of complexity and pushes up the cost of manufacturing. For example, the cost of a single dose of Novartis’ Kymriah CAR T-cell therapy was set at $475,000 back in 2017, which puts such a therapy out of reach for many who need them most. AI has a clear and transformative role to play in overcoming many of these challenges. Cell therapy potential Unlike traditional drugs, cells are living entities that can grow, migrate, target specific areas within the body and possess extremely complex functionalities. These characteristics make cells uniquely powerful tools for addressing complex diseases, particularly those that are difficult to treat with conventional methods, such as cancer, autoimmune disorders, and degenerative diseases. However, the intrinsic complexity of cells and cellular behaviour, and being able to understand how cells respond to different environments, how they proliferate or differentiate, and how they interact with each other, and how to harness those features to make robust therapeutic product can present significant challenges. This is where AI can offer immense benefits. Firstly, traditionally cell therapy research is time-consuming and expensive, often taking years of laboratory research and substantial financial investment to find a viable candidate. AI algorithms can speed up this process dramatically by analysing vast datasets, predicting cell behaviour, and identifying promising therapeutic approaches. By automating data analysis, AI can reduce the need for extensive manual work and can streamline the early stages of cell therapy development. In addition to this, the regulatory landscape for cell therapies requires extensive testing – often over many years. AI can track and organise massive amounts of data and can ensure traceability and compliance throughout the manufacturing process. Without AI, this can add another layer of complexity and cost, particularly when scaling up production. Developing cell therapy generates massive amounts of data and by leveraging AI, it could be possible to substantially derisk and reduce the time spent on regulatory submissions, testing and approval, and ultimately cut down on costs. Personalised treatments could also be advanced with AI. By analysing individual genetic profiles, medical histories, and molecular data, AI can help design highly targeted cell therapies tailored to specific patient characteristics. For example, AI could identify the most suitable cell types for a particular patient, predict how their immune system might respond to certain therapies, and determine the best conditions for cell differentiation, ensuring the therapy is not only effective but also minimally invasive. By tailoring cell therapies to the unique needs of each patient, it could be possible to reduce the likelihood of adverse reactions, and ensure that therapies are more likely to succeed and lead to long-lasting improvements. This precision approach could therefore significantly increase treatment efficacy and move beyond the traditional one-size-fits-all approach. Overcoming AI’s application barriers So, AI has huge potential, but in order to fully realise its benefits for cell therapy, there are technical challenges to address. While AI has been a hot topic for some time, integrating it into existing systems has proven complex. For companies not born as ‘AI-first’, adoption has been slower and more challenging than expected – AI cannot just be applied out of the box to data or augment existing processes and instantly deliver results. This applies in particular to large pharmaceutical companies, which can face a bigger challenge incorporating AI into their established processes; by contrast, start-ups and smaller biotech companies have more flexibility to leverage AI from the get go and build it into their workflows from the ground up. With the recent Nobel Prizes in Physics and Chemistry awarded for work in AI, it is clear that the shift towards AI and automation is growing, with an urgent need for this to be applied to impactful areas such as cell therapy. Companies must have the right infrastructure in place to unlock the power of AI. Advanced AI requires sophisticated computational resources, including high-performance computing systems and specialised machine learning-enabling hardware to manage the volume and complexity of cell therapy data ranging from imaging and genome sequencing to real-time tracking of cellular behaviour for analytics and process optimisation. Another key issue is the lack of high-quality, well-annotated data. Successful AI application relies on availability and access to large, information-rich, curated, organised datasets for effective algorithm training and deployment and to generate accurate predictions and insights. In the case of cell therapy, generating, obtaining and accessing such data is in only its infancy and remains limited and challenging. Most of the cell therapy-relevant datasets that are (publicly) available thus far are sparse, fragmented and/or difficult to integrate, making it difficult for AI systems to be effectively deployed. Furthermore, the transition from laboratory-scale research to large-scale clinical production also requires AI systems to adapt. Scaling involves changes in variables such as manufacturing environments, supply chain logistics, and even patient demographics, all of which can introduce new uncertainties. AI systems must be robust enough to manage these variations without sacrificing precision or reliability, which requires sophisticated algorithms capable of learning from and adapting to different conditions. Finally, when it comes to regulation, AI healthcare regulations require thorough validation of new technologies, and AI systems present unique issues in this area. The challenge is even bigger as the pace of regulatory updates may not always keep with AI’s rapid advancements. The ongoing development of these regulatory frameworks may create uncertainty for companies integrating these technologies. Last April, the UK’s MHRA released a white paper on strategic approach to AI vis à vis medical products, and in the US, last month the FDA issued the first draft guidance to support regulatory decisions on AI’s use to establish drugs and biological products’ safety, effectiveness and quality, an important step towards establishing and understanding how to support and regulate AI-enabled medical products. Both of these illustrate the active and pressing discussion taking place within regulatory agencies across the globe to both define a path ensuring that this new way of developing therapeutics meets robust scientific and regulatory standards while recognising and seeking to maximally embrace the potentially step-changing power of AI-enabled therapeutics. Shaping tomorrow The combination of AI and cell therapy is an exciting breakthrough area for science and healthcare. AI can help to organise data, make cell therapy development and production faster, and simplify and accelerate regulatory testing and approval, all of which will in turn help to reduce costs and improve efficiency delivering more therapies for more diseases and accessible to all. While challenges around scaling and access to data remain, AI’s ability to improve personalised treatments is clear. The goal of reducing human suffering and advancing medicine makes this effort not just scientifically important, but also deeply meaningful and of paramount importance. From DDW Volume 26 – Issue 2, Spring 2025 Read the digital issue About the author: Rafael Carazo Salas is Founder and CEO of CellVoyant, an AI-first biotechnology company with a mission to accelerate cell therapy development with the power of AI. He is also Professor and Chair of Biomedical Sciences at the School of Cellular and Molecular Medicine in the University of Bristol, formerly at the University of Cambridge.