Synthetic data has emerged as a transformative tool in biomedical and clinical research, offering solutions to critical challenges such as privacy, data scarcity, and bias. By creating artificial datasets that preserve the statistical properties of real-world data without revealing sensitive personal information, synthetic data allows researchers to work with valuable clinical information while maintaining strict privacy standards. This is particularly important in fields like healthcare, where data access is often restricted by stringent legal and ethical guidelines. Our lab has established the development and practical application of synthetic data methods, notably taking another big step forward through the use of large-scale projects like the UK Biobank.
Find out more about synthetic data and our lab’s work in our dedicated Research Pillar.
The UK Biobank, a comprehensive dataset containing health information from over half a million individuals, provides an invaluable resource for studying complex diseases such as lung cancer. However, access to this data is limited due to privacy concerns. By leveraging synthetic data, our team is pioneering ways to replicate the richness of this dataset while ensuring participant privacy is preserved. This allows researchers to conduct exploratory data analyses, develop predictive models, and test hypotheses without ever needing direct access to real, sensitive data.
Read our latest paper on Synthetic data for privacy-preserving clinical risk prediction, using a synthetic version of the UK Biobank, soon to be published in Nature Scientific Reports here.
Our lab’s cutting-edge research has shown that synthetic data generated by advanced models like DPGAN, ADSGAN and PATEGAN, used in this paper, can accurately mimic real-world datasets in terms of variable distribution and feature relationships. These synthetic datasets support a wide range of biomedical applications, from feature selection and model development to hyperparameter tuning. Our synthetic data models, trained on the UK Biobank data, have demonstrated success across multiple stages of clinical risk prediction, particularly in lung cancer prognosis.
Tom Callender Video
However, our lab’s work with the UK Biobank underscores only one facet of the potential of synthetic data to unlock insights from large-scale biomedical datasets without compromising privacy. In another paper, published in Nature Biotechnology, our lab and collaborators deliver a comprehensive review focusing on fairness, representation, and utility, we have shown how synthetic data can expand datasets, reduce bias, and ensure that underrepresented groups are adequately considered in research. Through the use of cutting-edge generative models, we are helping to push synthetic data beyond its traditional limits, opening new doors for biomedical research and collaboration.
Read the Nature Biotechnology paper here. (TBC)
(Quote by Dino)
In summary, our lab positions itself as a leader in in the development and application of synthetic data, driving forward innovations that will shape the future of healthcare and bioengineering. Follow us and our future research – or be part of it!
Interested in pushing the boundaries of Synthetic Data? We are looking for five outstanding machine learning PhD students to join our elite group at the University of Cambridge. Find out more here.









