Digital twins are virtual models that represent real-world systems in real-time. Their potential is vast, with their ability to analyse systems and provide predictive insights through machine learning. Digital twins work by enabling continuous monitoring and simulation of a physical object. Through live data integration and the use of AI, digital twins can identify patterns and run simulations to test different scenarios, including the fields of healthcare, pharma and biotech.

​

Within the medical field, a digital twin is defined as the ‘representation of a person which allows dynamic simulation of potential treatment strategy, monitoring and prediction of health trajectory, and early intervention and prevention, based on multi-scale modelling of multi-modal data such as clinical, genetic, molecular, environmental, and social factors’. Digital twins have been implemented in oncology, in the context of adaptive therapy: a cancer treatment strategy aiming to control tumour growth and resistance instead of fully eradicating it. In a real-time clinical trial, a patient’s tumour activity was inputted in a digital twin, monitored by prostate specific antigen (PSA) levels. The digital twin updated as new prostate specific antigen levels were inputted. When the antigen levels fell below 50%, treatment was paused, and when levels rose above this threshold, treatment was resumed. This adaptive method was successful in slowing tumour progression by keeping the number of resistant cells under control through competition with sensitive cells. Similarly, a retrospective simulation used mechanistic modelling to analyse PSA levels to study tumour growth. The digital twin took existing data on a patient’s blood tests and genetic markers, to visualise tumour growth and simulate hypothetical dosing scenarios. These two trials demonstrate how digital twins can be useful in both real-time and retrospective models to inform personalised cancer care.

​

Digital twins could transform the future of the drug discovery field. Traditionally, developing a drug is an expensive, lengthy journey, that requires years of laboratory work and clinical trials. On average, it can take about 10 years and billions of dollars to bring a single drug to market. Digital twins cold simulate biological processes and predict how a drug will behave in the body. These simulations allow scientists to test and identify drug candidates quickly. For example, companies like Siemens are working with drug manufacturers such as GSK to create digital replicas of laboratory and production processes, to reduce errors and optimise workflow. By virtually testing drug behaviour and production scenarios before moving to real-world trials, digital twins can shorten development timelines, lower costs, and ultimately help deliver effective treatments to patients sooner.

​

Clinical trials are costly and time-consuming. A developing strategy is an in-silico trial, which is a clinical trial that is conducted digitally through simulation and modelling. One of the biggest expenses in running a clinical trial is enrolling patients. This is significantly more difficult with rare diseases, where only a small number of patients are eligible to participate. Recruitment is also challenging when no effective standard treatment exists or when patients are reluctant to join a trial because they may be randomly assigned to receive a placebo or an existing treatment that may not help them. As a result, trials take longer to complete and may fail to enrol enough participants to produce meaningful results. Therefore, in-silico trials could be used to simulate both the control and efficacy sides by generating predictions of what an individual’s outcome would have been under general treatment versus different plans, allowing for more precise comparisons within a trial. This has been demonstrated in studies such as the VICTRE trial, which used nearly 3,000 virtual patients to compare breast imaging technologies, producing results that closely matched those of a conventional clinical trial. 

​

Digital twins in the biopharmaceutical sector offer representations of complex biological manufacturing systems. By integrating real-time data with hybrid models that combine mechanistic understanding and machine learning, they enable monitoring and predictive control of critical factors affecting biological processes, like pH, temperature, and nutrient dynamics. This facilitates improved optimisation of cell growth and product yield, as well as earlier detection of metabolic deviations that may lead to batch failure. However, the practical impact of digital twins remains constrained by several factors. Their accuracy is dependent on data quality and validity, which is challenging in biopharmaceutical systems characterised by biological variability and incomplete mechanistic understanding. Furthermore, the integration of digital twins with outdated manufacturing infrastructure presents significant technical barriers, limiting the current scalability of digital twins.

​

Whilst the future of digital twins seems promising, there are many challenges. One of them being the issue of data integration. Digital twins require the incorporation of multiple data types to accurately model processes and patient-specific pathways. However, healthcare data comes from many different sources, such as electronic health records, medical imaging, wearable devices, and genomic databases, and are often stored in incompatible formats across separate systems, making it difficult to exchange and synchronise data. The lack of standardised data formats further complicates the construction of reliable healthcare digital twins. As the use of digital twins expands, there is a growing need for a universal data platform that enables efficient and effective data exchange across healthcare systems.

​

Furthermore, creating digital twins to represent evolving conditions requires access to longitudinal data. This data is often scarce and may have missing data, which could hinder the creation of an accurate digital twin, especially in healthcare settings where not all patient data may be available. In addition, it is labour-intensive and subject to human error to generate properly labelling data, particularly in medical imaging or other diagnostic applications. Thus, maintaining data quality over time and across different sources can be challenging. Another potential issue is that digital twins could exacerbate healthcare inequalities through data bias. The accuracy of a digital twin requires a data model built on a balanced dataset. However, health data can be biased in various ways, such as being skewed towards certain demographics or conditions. Building digital twins based on biased datasets would worsen existing biases so it is imperative data is heavily monitored before being inputted into the technology. Additionally, it is imperative that patient data and privacy is secure and encrypted. Digital twins rely on large volumes of sensitive patient information, including clinical records, imaging data, genetic profiles, and data from wearable devices. This creates significant risks related to unauthorised access and data breaches. Robust data security measures such as encryption, access controls, and secure data storage are essential to ensure patient trust and ethical and responsible usage.

 

In conclusion, the potential of digital twins is very promising in many fields. However, their implementation on a large scale is only contingent on strict framework and regulation.

​

References:

https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00028-7/fulltext

https://www.nature.com/articles/s41746-024-01073-0

https://www.nature.com/articles/s41598-023-48747-5?fromPaywallRec=false

https://www.sciencedirect.com/science/article/pii/S2590156725000945#s0010