Guest Column | December 30, 2020

Seeing Double (For Good Reason): Using Digital Twins To Improve Pharma Processes

By Tim Sandle, Ph.D.

There are multiple reasons why engineers cannot visit a facility, not least the disruption wrought by 2020’s COVID-19 pandemic. Even when visits can be accommodated, there are time delays. The technological solution that has reverberated through industry, including pharmaceuticals, is digital twins. A digital twin is a virtual model of a process, product, or service.1 The pairing of the virtual and physical worlds not only enables systems to be adapted or repaired from afar, but the approach also further enables the analysis of data and monitoring of systems to avoid problems before they even occur. This can prevent downtime, foster new opportunities, and allow future-state solutions to be modeled by using digital simulations. This article looks at the digital twins concept and considers how the pharmaceuticals sector can take advantage of this particular crest of the digital transformation wave.

What Is A Digital Twin?

As applied to pharmaceuticals, the digital twin is a type of process assistant that has online and real-time access to data, including historical data and present process conditions. Real-time data is drawn from process analytical technology, quality data, and time series data. The digital twin can then be used to simulate the future based on all possible effective models. Modeling is often based on Bayesian analysis, a statistical paradigm that answers research questions about unknown parameters using probability statements. The origins date to the Reverend Thomas Bayes, who was possibly the first to apply conditional probability to provide an algorithm (referred to as “Proposition 9”) that uses evidence to calculate limits on an unknown parameter.2

Scientists often refer to Bayesian philosophy, an approach based on the idea that more may be known about a physical situation than is contained in the data from a single experiment. Hence, Bayesian methods can be used to combine results from different experiments, for example. From such analysis comes Bayesian decision-making. This involves basing decisions on the probability of a successful outcome, where this probability is informed by both prior information and new evidence the decision maker obtains.3

Digital twins can save pharmaceutical and healthcare manufacturers millions in expenditures across several aspects of the manufacturing process, including reducing the number of required process performance qualifications to setting a robust control strategy. Applications include:4

  • In process characterization studies, pharmaceutical companies set acceptance criteria and normal operating ranges using digital twins, well aligned with the FDA’s Quality by Design Initiative.
  • During process performance qualification (PPQ), biopharma companies calculate the necessary number of PPQ batches for validation, leveraging digital twin models to justify reduced PPQ run numbers.
  • Over the course of continued process verification (CVP), biopharma companies leverage digital twins to set alert limits, predict changes, and react to trends.

Digital Twins In Pharmaceutical Process Development

As indicated above, a digital twin is a virtual and connected model of a process, product, or service. Pharmaceutical manufacturers are making use of digital twins to drive process improvements throughout the manufacturing plant.

As an example of how digital twins can assist with process optimization and drug development, Atos, a digital transformation consultancy, and Siemens, the engineering firm, announced a project in 2020 designed to assist the pharmaceutical industry improve production through an innovation based on digital twins. The specific project involves creating a digital replica of the pharmaceutical production process (a process digital twin). The technological components that help drive the development include the Internet of Things (IoT), artificial intelligence, and advanced analytics. Each aspect has been put together to help provide improved efficiency and flexibility across the manufacturing of different pharmaceutical products.5

Benefits include reductions in experimentation time and waste. Furthermore, the use of digital representations has been configured to ensure experiments are of a constant quality and meet industry expectations around quality by design (which, in this context, means getting the product right the first time). For a sector reliant upon data, the solution is also capable of using collected data to optimize process quality and machine reliability.

The end product is a complete virtual replica of each step within the manufacturing process. The digital replica connects with IoT sensors installed within the physical plant. This is an example of “smart pharma,” where the orchestrated interaction between online sensors and digital models leads to enhanced control of the production process. The process sensors collect real-time data, which can be drilled down in order to provide a real-time assessment of operations. Improvements can be driven through a comparison of digital representations of the process and the physical reality of the process. Where the physical reality does not match the digital model expectations, improvements can be sought. In addition, the digital model can be used to assess when and where the physical process may go awry; this is something that can be achieved through predictive models that can analyze the real-time data.

In terms of product development, the Atos-Siemens platform has been used to create simulation of chemical mixing processes. This enables scientists to run multiple variants of different processes, with the end result of ensuring that the optimal mix of chemicals is used in live production. The net effect is to reduce waste and, hence, save costs, as well as facilitate getting new medicines to the market more quickly.

The biotechnological process is heavily reliant upon model-based methods. This is due to the process itself being dependent upon prediction and control capabilities. Delays and false starts often arise due to the lack of suitable models. In a different example, by using digital representations, TU Wein researchers demonstrated how a target-oriented mechanistic process model can be developed for biochemical-based drug development. This was based on designing a simplex algorithm that helped with bioprocess development and optimization. A similar data-reliant model from Johannes Kepler University involved constructing the architecture necessary for performing computational model life-cycle management, which resulted in an end product providing the infrastructure needed for computing predictions in real time.

Using Digital Twins To Reduce And Address Out-of-specification Results

The use of digital twins can help to reduce the number of out-of-specification (OOS) events and assist with tackling the issues that stem from them. Some examples are:1

  • Better specifications, fewer out of specification events: The simulations will predict bulk substrate and finished product specification variations more reliably than individual unit operation models.4 With these simulations, precise specifications and control limits can be estimated not only for the intermediates but for the end results. This will result in fewer deviations and OOSs.
  • Seamlessly predicting the future based on past data and present simulations: The historical data, the present data, and the models combine within the simulation to give the most precise prediction possible, given the complexity of bioprocesses.
  • Continuously improving: Machine learning algorithms can be easily incorporated to take in new production data and improve the models, thus allowing for increasingly precise models over time.
  • Faster problem solving: Atypical outcomes can be traced back through the simulations to investigate potential special cause variation, whether univariate or multivariate.

Digital Twins And Drug Development

The application of digital twins in drug development has been shown with a second example from Atos, relating to the path toward a vaccine for the novel coronavirus SARS-CoV-2. It follows that once a vaccine has been proven, the challenges around mass production can be as complex as developing the actual vaccine in the first place. Such pressures are especially acute when there is an imperative to accelerate the time to get the drug product to market.6

The time to deliver a drug to market can be reduced through the collection and analysis of data. Big Data analytics, together with advances in computing, can not only help with the core development of a drug product (as with formulation and disease modelling run as overlapping simulations), but they can also assist in streamlining the pharmaceutical production process. Digital twins can help foster improvements in manufacturing and take a developmental drug to scale. The process of constructing a digital replica starts with having each manufacturing stage equipped with in-line sensors. These sensors enable various sets of data to be collected and interpreted, in actual processing time. When the analyzed data is combined with physical, chemical, and biological data, models can be produced and these models provide the foundation to the digital twin.

The developed digital twin becomes a real-time replica of the physical production processes, enabling pharmaceutical manufacturers to optimize each step and to introduce required changes to improve the process or to create simulations to predict what a particular modification will do to the next stage or to the finished product. The success of predictive improvements grows through machine learning algorithms. This approach also assists with making more general process improvements through gaining more accurate assessments of changes prior to implementing change control processes. This method is in keeping with the quality by design paradigm.6

Digital Twins And Biopharma Logistics

Global competition is driving change in the logistics sector, and to remain competitive, logistics companies need to offer customers security and speed and to remain cost effective. Customers are also seeking traceability and transparency from logistics solutions providers. One way to meet these expectations is for pharmaceutical sector connected logistics firms to offer digital twins, which allow logistics providers to provide their partners complete visibility into a product’s life cycle. In addition, logistics companies can use the design element from digital twins projects to increase operational efficiency. This technology can help logistics providers make better-informed decisions for optimizing material flows and monitoring supply chain processes. Examples of applications include supporting activities like the optimal deployment of a forklift within a warehouse, the use of robotics to assist with sorting materials, and optimizing the use of delivery trucks.

Digital twins can assist with assessing events that might occur during transportation (such as damage or temperature excursions, factors that impinge on good distribution practices). The digital twins process also enables real-time updates to be gathered on the status of goods and assets throughout the supply chain. For the logistics provider, this can mean faster responsiveness, as well as opportunities to reduce waste, achieve improved inventory control, and optimize the use of warehouse space. Complexities involved with maximizing the advantages of digital twins require bringing together cloud computing, artificial intelligence, and advanced visualization tools to assist with the digital twins process.7


When implemented well, digital twins provide an array of advantages for pharmaceutical manufacturers, ranging from process development to logistics. Not only can software and sensors be used to create living digital simulation models, but such models can also update and change as their physical counterparts change, providing real-time status and working conditions.

This article has been adapted from chapter 6 of the book Digital Transformation and Regulatory Considerations for Biopharmaceutical and Healthcare Manufacturers, Volume 1, written by Tim Sandle and co-published by PDA and DHI. Copyright 2021. All rights reserved.


  1. Taylor, C. (2018) What is bioprocessing digital twin?, Exputec, at:
  2. Bayes, R. and Price, M. (1763) An Essay towards Solving a Problem in the Doctrine of Chances. By the Late Rev. Mr. Bayes, F. R. S. Communicated by Mr. Price, in a Letter to John Canton, A. M. F. R. S, Philosophical Transactions of the Royal Society of London. 53: 370–418. doi:10.1098/rstl.1763.0053
  3. Zahel, T. (2018) Bayesian bioprocess data analytics in R&D and GMP manufacturing, Exuptec, at:
  4. Taylor, C. and Herwig, C. (2019) Integrated Process Modelling – Very Useful Bioprocess Digital Twin, IPSE, at:
  5. Atos (2020) Atos and Siemens introduce Digital Twin solution within the global pharmaceutical industry, at:
  6. Thomsen, N. (2020) The role of digital twins in producing a COVID-19 vaccine, Atos, at:
  7. Sandle, T. (2019) Digital twins: Redesigning logistics, creating new efficiencies, Digital Journal, 10th July 2019, at:

About The Author:

Tim SandleTim Sandle, Ph.D., is a pharmaceutical professional with wide experience in microbiology and quality assurance. He is the author of over 30 books relating to pharmaceuticals, healthcare, and life sciences, as well as over 170 peer-reviewed papers and some 500 technical articles. Sandle has presented at over 200 events and he currently works at Bio Products Laboratory Ltd. (BPL), and he is a visiting professor at the University of Manchester and University College London, as well as a consultant to the pharmaceutical industry. Visit his microbiology website at