Automated production of mRNA therapeutics

The price of a drug is not only dependent on which raw materials are used but is also determined by the production process. High prices are charged for some cancer drugs, as they can only be produced in small quantities at great expense. Keyword: mRNA therapeutics. This is precisely where employees from seven Fraunhofer Institutes want to start in the lead project, RNAuto. We spoke to Rolf Hendrik van Lengen, who is involved in the project on behalf of the Fraunhofer Institute for Experimental Software Engineering IESE.

Mr. van Lengen, what goal are you pursuing in the lead Fraunhofer project, RNAuto, in which the Fraunhofer IESE is also involved?

Rolf Hendrik van Lengen: In RNAuto, the Fraunhofer Institutes IESE, IMM, IMS, IPA, IPT, ITEM, and IZI are working together on the automated production of individual mRNA therapeutics, for example for cancer therapy. In the future, it should be possible to produce these cost-effectively in larger quantities to make them more affordable. Until now, production has been carried out manually, which we want to change. This is because it means that these drugs are very expensive to produce for an individual patient. A cancer drug can currently cost up to €235,000, which is unacceptable in the medium and long term. Our vision is therefore to use our approach to reduce production costs to a tenth of the current price.

How do you intend to achieve the goal of an automated production plant?

van Lengen: We want to build systems that are made up of scalable, flexible and manufacturer-independent production modules. These components can be replaced via plug and play in the event of a defect. The test system currently measures 1.20 m x 0.90 m x 0.90 m.

We also want to simplify quality control and digitalize it. This means that the quality of the product or products is continuously measured during the production process. Up to now, quality control in the industry has only taken place at the end of a batch. There is a high risk that an entire batch may have to be disposed of if something has not been produced according to specifications. Or, as is also done in practice, random checks are carried out in order to react accordingly in the event of defects.

This is acceptable if you have a 500-liter bioreactor filled with vaccine, where it is not very noticeable if a partial quantity is removed. But with mRNA production it is a problem because you only produce very small quantities. If you were to remove some of this quantity after each sub-step, you would end up with – to put it bluntly – nothing left over.

We therefore want to develop a screening system with digital process control that can carry out quality control during production without destroying the substance produced. This data-driven online quality control could then be used to encapsulate the mRNA in lipid nano transporters.

How do you check whether the system is mixing the medicines as desired?

van Lengen: We check fluctuations in the viscosity of the mixture, refractive indices, conductivity, temperature and pH value. We also work with a digital twin that maps the processes and product quality. Naturally, high regulatory requirements are placed on the production facilities for a pharmaceutical product. These must be complied with, and the digital twin also checks this. To do this, it receives the relevant values from the system's sensors. The next step is to actively intervene if, for example, the temperature falls below a certain parameter. This should then automatically increase again, or a warning should be issued if this is not possible, perhaps because a unit is defective and needs to be replaced.

What role does artificial intelligence (AI) play in this process?

van Lengen: An AI can help at many points during production. For example, it can help to rule out an active ingredient candidate as useless from the outset or it can monitor and improve the production of a plant. The quality assurance of our plant and the process control, i.e. the control of pumps, mixers etc., are mapped completely digitally – via a digital twin and also with the help of AI-supported software tools.

The AI supports us by reporting, for example, that a parameter is not set optimally. At least that is the plan for the end of the project. However, we are currently still working on the system and testing whether it does what it is supposed to do, namely package mRNA. For cost reasons, we are currently testing with substituents. Towards the end of the project, we will test with a vaccine candidate against the West Nile virus, as a project partner has a patent on it. We have an mRNA that is coded for this virus. We are looking to see if our plant can package it correctly and make it effective.

What software do you use for production monitoring?

van Lengen: We use various software tools. The COPE process control software from the Fraunhofer IPT, which controls the individual components. This means, for example, that the software gives a pump the command to move in a certain direction in order to dispense liquid. We also use the Industry 4.0 middleware Eclipse BaSyx from Fraunhofer IESE, which was actually developed for Industry 4.0 to control production processes and quality control.

We combine all of this to be able to produce quickly and in a targeted manner with the screening system. We can digitally model risk management, i.e. represent it as an abstract model. Our colleagues in the department have developed algorithms based on this abstract model that automatically generate digital twins for quality control. This has the advantage of being able to track changes to regulations at lightning speed.

In these cases, it is built into the model, a new twin is generated and the information is transferred, allowing you to react very flexibly. This also applies to individual parts, for example if a pump needs to be replaced. If the installed model is no longer available, for example, I can easily use a different manufacturer. The only challenge is that every change to the system must be recertified. This is because it produces medicines that are to be taken by people. Special regulations apply here, as the quality must be consistently high. We are currently looking for digital methods of certification to be able to certify faster than has been possible to date. Otherwise, you would immediately lose the advantage you have gained by quickly replacing a component. This aspect will be the subject of further research over the next one to two years.

When is the project due to be completed?

van Lengen: By the end of 2025, the system should be completed at the Fraunhofer IMM in Mainz and also made available to industrial partners.

Looking very far into the future: When do you think patients will benefit from the new procedure?

van Lengen: Good question, I can only speculate. We are thinking about designing, developing, certifying and setting up such systems in cooperation with other institutes. Production could then be set up at a university hospital, for example. But realistically, it will probably take around five to ten years to get to that point.