Cell and gene therapies are exciting new treatment paradigms that offer curative, or potentially curative, treatments in a variety of indications, many of which are resistant to traditional approaches. However, the uptake of advanced therapies has been limited by their high cost and consequent poor patient access. One of the most significant factors causing the high cost is the difficulty in manufacturing the treatments. Here at Cambridge Consultants, we've been looking at how artificial intelligence (AI) can play a role in reducing the cost of manufacture of cell and gene therapies and hence make these treatments available to more patients who so desperately need them.

There are several areas where artificial intelligence could play a role in improving the efficiency of cell and gene therapy manufacture. These include R&D, product manufacture, patient stratification and optimising workflow and logistics. Let's look at each of these in turn.

There have been a number of groups using AI in both classical drug development for small molecules and biologics, and also for things like design of CAR constructs in CAR-T therapy. On the gene therapy front, it's also possible to use AI to design improved versions of enzymes that can help to overcome metabolic defects in patients. Using AI in this way for protein design to improve enzyme efficiency has a long track record already in industrial biotech, but is only just now starting to be applied to therapeutic development.

Another important aspect where AI has started to pay off has been in the analysis or even design of experiments in the drug development process, allowing DOE spaces to be explored to develop the most effective therapeutic entities.

Autologous cell therapies

Product manufacture is a well-known issue, particularly in cell therapies where in some cases a patient’s own cells are being used as the starting material to make the therapeutic product, which means there is a batch size of just one product. In these autologous cell therapies, the input material – the white cells collected from the patient – can vary significantly in its quality, proliferative potential and activity. It's very difficult for manufacturing processes to be able to cope with this lack of definition around the specification for the input material.

We see that a big contributing factor to the cost of cell therapy manufacture are batch failures, which have many different causes, but certainly this variability of input material is a major one. Using AI to characterise these input materials to better predict how the cells will grow during the manufacturing process, and to identify batches which won't meet product release specifications early in the manufacturing process before the full expense of manufacture has been incurred, is a very attractive prospect.

Another exciting aspect of AI in this setting is being able to bring together large data sets from the manufacturing processes of different cell types from different patients on different hardware platforms. This gives understanding and insights that are more generally applicable across a range of therapeutic approaches, that will in turn lead to better results when growing cells in a bioreactor to make the therapeutic product. This data could lead to AI-driven, real-time protocol modifications without user intervention – reducing costly labour and achieving more automated manufacturing processes. We are seeing bioreactor companies increasingly include AI as part of their process controls for this very reason.

Of course, the variability from one patient to the next doesn't just affect the success of manufacturing. Different patients have different severities of disease, and this has a big impact on the likelihood of successful treatment with an advanced therapy. AI models for understanding diagnostic data and how that can be applied to patient stratification and hence selection, for example in clinical trials, is likely to be a powerful tool. It will help us more rapidly understand in which patient populations an expensive new cell and gene therapy treatment should be used.

DNA sequencing technologies

This can have benefits both for therapy developers who are better able to demonstrate the therapeutic potential of their product, and also for payers who can be more confident that the treatment will have a beneficial effect in the particular patient sub-group. Other complementary technologies such as DNA sequencing technologies and metabolomics are helping to provide the data sets that will enable these AI models to be able to better differentiate patient subgroups from each other.

Finally, the process of manufacturing cell gene therapies, pre-treating the patient appropriately, administering the therapy in the right care setting and then following up appropriately is a very complex and difficult to schedule workflow. AI models are again able to help coordinate the scheduling of different doses and different patients across different care centres. In order for the industry to be able to successfully move from treating a few hundred patients a year into treating many tens of thousands, this will be essential as constraints around manufacturing facilities are likely to remain a bottleneck in the delivery of these treatments.

AI models can be used to ensure the most efficient utilisation of manufacturing facilities which will again reduce costs and improve outcomes. This also goes hand in hand with overall digital transformation of healthcare delivery in different settings.

We're very fortunate that at CC we have a wide variety of experts from different fields such as cell biology, process engineering, software and artificial intelligence. It means we are able to work together to help our clients in the cell and gene therapy space identify the best solutions for their challenges and to help get their therapies onto the market and into the clinic to help patients as efficiently as possible. Cell and gene therapies remain an extremely exciting approach to the treatment of a range of diseases and we're excited to be able to play our part in ensuring as many patients as possible can benefit from these technologies.

Author
James Hallinan
Head of Business Development, Bioinnovation

James specialises in bioinnovation and early-stage technology commercialisation across life sciences and healthcare

Connect with James