Umetrics Suite Blog

Digital Twins Offer a Data-Based Approach to More Effective Biopharma Process Control

December 11, 2019

For pharmaceutical companies facing challenges such as rising costs, sterner regulations and declining profit margins, innovative new technologies like artificial intelligence (AI) and digital twins have become part of an essential strategy to future-proof their businesses. A digital twin is the next evolution of machine learning combining advanced data analytics and equipment simulation with comprehensive system models that blend historical information with real-time data to predict the future of a process. According to Gartner, the digital twin concept was one of the top 10 strategic technology trends in 2019.

digital twin in pharmaceutical manufacturing process

Digital twins are evolving to be the next big thing in pharma and can help reduce production costs by keeping processes in control and speeding up development.

So what exactly is a digital twin?

A digital twin is a virtual representation of a production process based on historical and current data. With seamless transmission of data between the actual physical bioprocess and the virtual entity, the virtual twin is able to create a simultaneous replica of an existing real-world process in order to predict the outcome in advance.

A digital twin is the next evolution of digital modeling using virtual simulation combined with process control. For pharma companies, perhaps the most exciting benefit of a digital twin is being able to simulate a process variation and understand the resulting affects, without having to spend months or years in scale production or testing.

To be effective and control a process efficiently, a digital twin must react very fast. That means data must be captured in real time and be part of a continuous feedback loop. Fast simulations are needed to predict the current state, and the operator or system (if automated) must be able to act quickly to keep the process in the ideal range of performance attributes.

The digital twin runs in real-time or in batch mode and incorporates development as well as historical manufacturing data. It uses models trained on PAT data, quality data, and time-series SCADA data, using advanced data analytics methods like those found in SIMCA software. Digital twins can monitor and predict critical quality attributes and key performance indicators throughout the process chain.

The evolution of process models toward digital twins

In one sense, advanced multivariate data analysis (MVDA) is a step toward creating a digital twin, especially when running real-time process control. The evolution to digital twin from real-time process monitoring means having a complete replica of a biological organism or process. You start with knowledge of the process in order to facilitate in silco development (models created digitally). This moves across four levels:

  • Equipment simulation. Digital replicates fed by data from historical and ongoing process models.
  • White box model. A model that uses predictions to determine what variables clearly have influence. These are important to train the model as part of the machine learning process.
  • Online monitoring. Using online release or real-time release, you’re managing quality by design with testing, but feeding data back into the process.
  • Self-running bioprocess. Using adaptive control lets your digital twin push your process data back to your batch to make corrections automatically when deviations occur.

Why use digital twins?

Digital twins offer a number of benefits in biopharma process development and pharma manufacturing. From the earliest phase of development through the full-scale manufacturing process, digital wins provide enormous potential for cost savings, error reduction and process control. They can help:

1) Reduce development costs

From vaccine development to bioprocess monitoring, digital twins offer companies the ability to simulate a process variation and understand the resulting affects, without having to perform a single run at scale. A digital twin can explore the early design space, suggest further experimental designs, and then train on the new data as it goes along. This grows your system knowledge and the speed at which new idea are developed, explored and ready for production exponentially.

2) Support more sustainable development

With access to more precise data and compelling analytics, you’ll be able to better control your overall drug manufacturing process. The result is less waste of time, resources and materials.

3) Enable more effective process control

Digital twins create real-time simulations of production processes that enable early intervention to correct deviations before they impact the outcomes of batches or affect downstream processing. Early deduction and even prediction of processes going astray are a key benefit of digital twins, and also one of the core principals of real-time process control.

4) Allow faster development and testing of candidates

By predicting the outcome of processes, digital twins help reduce the number of physical real-world tests or process completions needed to discover or validate new drug candidates. Digital twins enable process engineers to visualize the operation, simulate and optimize parameters, perfect the batch design, and make assisted decisions.

What is needed to create a digital twin?

Bioprocess digital twins leverage data from process development, design of experiments, manufacturing, mechanistic modeling, and even risk assessments. Therefore, a scalable, asset-centric data foundation for both manufacturing and development data is critical. To build digital bioprocess twins, asset connectivity is required, as are statistical analytics and mechanistic modeling, all in one environment.

For example, to build a digital twin prototype for the production of adjuvants by microfluidics¹, you would need:

  • a convincing prototype
  • efficient communication between online sensors and hybrid models
  • a process feedback-loop that can support adaptive control
  • a 3-d digital copy of the process with real-time visualization of process parameters and critical quality attributes.

Read more: Design of Experiments enables the optimization of transfection efficiency in line with QbD principles

Steps in building a digital twin

  • Define the needs and functionalities 

  • Develop PAT 

  • Design and implement the IT architecture 

  • Create, train and validate the models 

  • Ensure real time data processing and data integrity 

  • Build the automation 

  • Design the user interface 

  • Document 


Digital twin technology is evolving to become the next big thing in pharma. With digital twins in place, pharma manufacturers will have a solid digital footprint of their products throughout the development cycle, starting from the design phase, all the way to distribution.

Read these related articles:

How digital transformation helped Amgen implement real-time process control

Which data analytics methods best support a quality by design approach for biosimilars?

Find out more

Find out more about how Amgen is using data analytics and digital transformation technologies to improve processes and predict problems before they occur.

Get Amgen Presentation




1. Sandrine Dessoy, Christos Varsakelis (GSK), "Digital twin for a vaccine process,” presentation at 2019 PDA Europe, 3-4 Sept, 2019, Munich Germany.


Topics: Real Time Process Monitoring, Quality by Design (QbD), Pharmaceutical manufacturing

Kai Touw

Written by Kai Touw

Kai Touw is a (Bio)Pharma Market Manager at Sartorius Stedim Data Analytics. He is a driven and enthusiastic technology evangelist bridging the world of data science into pharmaceutical and biopharmaceutical processing.

Search the Blog

    Subscribe to the Blog

    View the:

    Data Analytics Glossary of Terms

    List of Webinars

    Get a free trial