Accelerating Vaccine Development With Automation
Automation and artificial intelligence have the potential to tackle many of the current barriers to rapid vaccine responses.
Complete the form below to unlock access to ALL audio articles.
Typically, a new vaccine will take 10–15 years to develop, in a long, complex and extremely expensive process.1 However, when the COVID-19 pandemic brought the world grinding to a halt in March 2020, the need for rapid vaccine responses to emerging pathogens came to global attention. Subsequently, thanks to worldwide collaborations, prioritized workflows and years of related coronavirus research, the Pfizer BioNTech mRNA vaccine was fully tested and approved for public use less than a year later.
Despite the successes of the COVID-19 vaccines, there is still a multitude of diseases lacking efficient protection, and future emerging pathogens remain a constant threat. As current long development and manufacturing pipelines continue to put vulnerable populations at risk, improvements to almost every aspect of vaccine production are required, from vaccine design to product manufacture, while maintaining high safety and efficacy standards.
Automation and artificial intelligence (AI) have the potential to tackle many of the current barriers to rapid vaccine responses at almost every stage of research, production and distribution. A wide range of automated and computational approaches are being explored and implemented in the field.
Improving the success of vaccine design
One of the biggest challenges in vaccine development is simply selecting a target antigen. A successful target antigen must be highly conserved to avoid mutation and vaccine escape, but immunogenic enough to trigger a robust immune response and long-term memory. For rapidly mutating pathogens such as SARS-CoV-2, or complex pathogens lacking obvious antigens such as Mycoplasma tuberculosis, finding a successful vaccine target can be highly challenging.2 Computational approaches use genetic data, bioinformatic tools and deep neural networks to predict potential antigen targets and identify vaccine candidates for further testing.3 Taking this principle a step further is the idea of designing artificial antigens to induce the best possible immune response.
“Pathogens are moving targets, they’re constantly evolving to evade the immune system, so if a vaccine targets a very specific epitope, it’s possible that the pathogen will evolve to evade that,” Dr. Joshua Blight, Co-founder of vaccine design startup Baseimmune, explains. Baseimmune’s research aims to use a novel machine-learning platform to design artificial antigen targets, drawn from the entire pathogen and all its possible variants. These antigens can then be inserted into any vaccine delivery platform, such as RNA or viral vectors, therefore enabling a vastly accelerated vaccine design process.
“Ideally, we need to stop chasing variants, and start predicting them,” Blight says. “In the early days of the COVID-19 pandemic, our computational platform predicted 80% of mutations in the virus that have now been found in common circulating variants. We’ve been able to look forward and see what the virus could become and use that information, along with data on current variants and related viruses, to design a universal coronavirus vaccine.” In addition to the universal coronavirus vaccine, this technology has also produced vaccine candidates for African swine fever and malaria, all of which are currently entering preclinical phases. If successful, artificial antigens could lead to a new level of pandemic preparedness, whereby automated, AI-based programs could generate new vaccine designs in response to, or even in advance of, emerging diseases.
Increasing the accuracy of trial results and reporting
Vaccine development generates enormous amounts of critical and sensitive data, especially during clinical trials. Collecting, analyzing and reporting such large amounts of data is incredibly time consuming, and requires high levels of precision, as even small errors can lead to trials being halted. Automating repetitive and error-prone tasks can increase efficiency and improve data integrity, and many research centers are adopting electronic laboratory notebooks in order to support this.4
One of the most important aspects of clinical trials is the reporting of adverse events (AEs). Electronic AE reporting systems such as the Vaccine Adverse Event Reporting System (VAERS) are already well established but are not generally capable of circulating or analyzing data. Automated AE data management systems can collate and analyze data, then immediately forward the information to key individuals, such as investigators, sponsors and regulatory bodies in compliance with good clinical practice (GCP) guidelines. This is particularly useful in large multi-center trials where data is generated in multiple locations.5 Automation can also be used to facilitate the writing of clinical study reports (CSRs). CSRs are essential for approval of new vaccines but require large amounts of data handling. Large data sets can be prone to error when input manually, and so benefit from automated data handling and checking programs, to avoid repeated rounds of review.6 Automation across clinical trials can support faster study completion, reduce costs and help to streamline approvals processes, which can be critical in pandemic and outbreak scenarios.
Supporting continuous process development
The development of new vaccine platforms, such as mRNA, has opened up new opportunities for automated workflows. Most traditional vaccine types, such as inactivated and subunit vaccines are cell-dependent and are produced by batch manufacture. These methods can be inefficient, and difficult to implement on a large scale. In comparison, mRNA manufacturing is a cell-free, enzymatic process, resulting in faster turnarounds, more consistent products and lower risks of contamination.7
Dr. Zoltán Kis, lecturer in the Department of Chemical and Biological Engineering at The University of Sheffield is developing continuous process methods and computational models for RNA vaccine production. The goals of this research are to increase manufacturing yield, ensure consistent high quality, reduce manufacturing costs and improve distribution efficiencies. “A continuous process requires the continuous insertion of raw materials and continuous purification. This means that everything is operating consistently at optimal conditions, and it’s much more efficient than batch production,” Kis says. “However, if you implement continuous processes, you need to have continuous monitoring too – of both process parameters and quality attributes, to avoid accumulating bad product.”
One of the challenges Kis’ research needs to overcome, is the fact that many of the critical quality attributes (CQAs) monitored in mRNA vaccines, such as RNA integrity, cannot be assessed in real-time.8 Measuring RNA degradation for example, consists of collecting a sample of product, then analyzing it using a technique such as high-performance liquid chromatography (HPLC), which can take up to an hour to produce a result. This is not fast enough for continuous processes. “We’re developing a computer model called a software sensor, which can calculate CQAs based on information that can be measured automatically and continuously in process, such as pH, temperature and product concentrations. Since this is constant, you can create profiles of the CQAs over time” says Kis. “Then, the next step would be to produce a model that can use these profiles to predict how the CQAs will change a few minutes into the future. If we couple this with automated feedback controls, we could have a system that could predict, say, RNA degradation from the current trajectory, and automatically fix it before it becomes a problem.”
Increasing productivity in manufacturing
The COVID-19 pandemic resulted in extremely high demand for vaccine production, leading to many manufacturers looking to intensify their processes and productivity. While there are multiple avenues to achieve this, the overall end goals are to increase yield and product quality, while reducing time and cost.9 Automation-enabled manufacturing facilities are one example of current process intensification.
William Whitford, is a strategic solutions leader in the life sciences division of DPS Group, a consulting, engineering and construction management firm specializing in high-tech buildings, which has supported vaccine manufacturing facilities to enable the COVID-19 vaccine production efforts of such sponsors as Pfizer and Moderna. “Automating manufacturing can be an expensive and complex endeavor,” says Whitford. “When addressing the automation of a process, the exact machinery to fit that need might not exist, and you need to modify either the process, or the instruments. Then, once you have your automation in place, it needs to meet such regulatory requirements as validation and qualification. There are a lot of things to consider.”
Prefabricated manufacturing elements and modular components are becoming increasingly popular as a way of tackling these challenges. “A comprehensive application of such an approach is represented in the so-called factory-in-a-box,” Whitford says. “These are rapidly deployable, remotely managed, modular manufacturing facilities that support decentralized production. Vaccines can be produced to meet demand in local areas, without the need for long supply chains.” The flexibility of these systems can provide automation-enhancement and streamlined production for a range of applications − meaning they can be easily adapted to respond as demands for different vaccines vary. Meanwhile, their prefabricated and modular nature provide fast deployment of current good manufacturing practice (cGMP) compliant environments.10
Innovating distribution and surveillance
The potential roles for automation in vaccine production don’t end with manufacturing. Once produced, vaccines must reach their destination with full potency intact, in order to ensure sufficient immunity in vaccinated individuals. In the case of inherently unstable vaccine platforms such as mRNA, this means distribution and storage under strict cold and ultra-cold conditions. However, these cold chains can fail, resulting in ineffective, wasted vaccines, damage to public trust and potentially unprotected recipients.11 Another aspect of Dr. Kis’ research is developing automated stability modeling for vaccines throughout the distribution chain.12
“It isn’t fully understood just how much CQAs and shelf life can be impacted by temperature fluctuations during the distribution of vaccines,” says Kis. “We’re working on developing a model that can take temperature and time readings from vaccine shipments and use those to calculate changes to CQAs and the subsequent remaining shelf life, specific to each individual shipment.” A model like this could be particularly useful in low- and middle-income countries, where limited infrastructure can increase the chance of cold chain failure.
Following licensure and administration of a vaccine, surveillance of safety data is an ongoing process. As of January 2023, approximately 70% of the world’s population has received at least 1 dose of a COVID-19 vaccine, with over 13 billion doses administered globally. Worldwide vaccination programs such as COVID-19 generate enormous amounts of data, which, without adequate analysis and visualization frameworks in place, could be sorely underutilized. Use of automated data visualization dashboards, such as the Vaccine Safety Datalink (VSD) COVID-19 Vaccine Dashboard, can concisely summarize vaccine safety reports, and aid in the rapid review of analyzed data.13 This ensures that the information can be used correctly to inform policy makers’ key vaccine safety decisions, and support public health vaccine initiatives.
Is the future of vaccine production fully automated?
Automation and AI can be a powerful tool in the development and manufacture of vaccines. It can increase the speed and success of developing novel vaccines and intensify quality product yields in the manufacturing process. Continuous technological developments may see even more automation throughout the vaccine production and distribution chain. “Eventually, as with other industries, automated robotics will carry out multiple tasks that fulfil defined requirements. Flexible and adaptable autonomous robots are demonstrating their ability to perform many of the repetitious tasks in both the development lab as well as operations on the manufacturing floor,” says Whitford.
However, there are still considerable areas of vaccine development and manufacture where humans provide the advantage over automation. “Robotics and automation are very powerful when it comes to defined procedures, but humans have far more latitude. An AI-enabled automated process might be able to solve issues within the scope of its own activity, but it is not aware of what is going on around it. Humans can especially accommodate assignable cause variation, whether that’s in the process or the environment, and that’s a big hurdle for automation,” concludes Whitford.
References
1. Vaccine development, testing and regulation. The College of Physicians of Philadelphia. https://historyofvaccines.org/vaccines-101/how-are-vaccines-made/vaccine-development-testing-and-regulation. Updated May 7, 2022.
2. Owens J. Faster approaches to vaccine design. Technology Networks. https://www.technologynetworks.com/tn/lists/faster-approaches-to-vaccine-design-362921. Published June 22, 2022.
3. Lv H, Shi L, Berkenpas JW, et al. Application of artificial intelligence and machine learning for COVID-19 drug discovery and vaccine design. Brief Bioinform. 2021;22(6). doi: 10.1093/bib/bbab320
4. Kwok R. How to pick an electronic laboratory notebook. Nature News. https://www.nature.com/articles/d41586-018-05895-3. Published August 6, 2018.
5. Richesson RL, Malloy JF, Paulus K, Cuthbertson D, Krischer JP. An automated standardized system for managing adverse events in clinical research networks. Drug Saf. 2008;31(10):807-822. doi: 10.2165/00002018-200831100-00001
6. Hong MK, Yao HH, Pedersen JS, et al. Error rates in a clinical data repository: lessons from the transition to electronic data transfer—a descriptive study. BMJ Open. 2013;3(5). doi: 10.1136/bmjopen-2012-002406
7. Rosa SS, Prazeres DMF, Azevedo AM, Marques MPC. mRNA vaccines manufacturing: challenges and bottlenecks. Vaccine. 2021;39(16):2190-2200. doi: 10.1016/j.vaccine.2021.03.038
8. van de Berg D, Kis Z, Behmer CF, et al. Quality by design modelling to support rapid RNA vaccine production against emerging infectious diseases. NPJ Vaccines. 2021;6(1). doi: 10.1038/s41541-021-00322-7
9. Whitford W, Sourabiè AM, Varshney DB. Enhancement of cell-based vaccine manufacturing through process intensification. PDA J Pharm Sci Tech. doi: 10.5731/pdajpst.2020.012583
10. Whitford B. Bioprocess intensification: aspirations and achievements. BioTechniques. 2020;69(2):84-87. doi: 10.2144/btn-2020-0072
11. Rogers B, Dennison K, Adepoju N, Dowd S, Uedoi K. Vaccine cold chain. AAOHN J. 2010;58(9):337-346. doi: 10.1177/216507991005800905
12. Kis Z. Stability modelling of mRNA vaccine quality based on temperature monitoring throughout the distribution chain. Pharmaceutics. 2022;14(2):430. doi: 10.3390/pharmaceutics14020430
13. Kenigsberg TYA, Hause AM, McNeil MM, et al. Dashboard development for near real-time visualization of COVID-19 vaccine safety surveillance data in the vaccine safety datalink. Vaccine. 2022;40(22):3064-3071. doi: 10.1016/j.vaccine.2022.04.010