- adsilico, a University of Leeds spinout, has raised £10M in a Series A round led by PXN Ventures.
- Women made up 33% of participants in 195 high-risk implant trials.
- The funds will extend its heart model portfolio to cover neurovascular devices.
Heart valves and stents typically get to market following trials with only a few hundred patients in them, and women are often under-represented. adsilico wants to fix that by using generative AI to create virtual hearts and blood vessels from anonymised scans.
The company has raised £10M in a Series A to enable this work. PXN Ventures led the round, with Northern Gritstone and Parkwalk also joining in. Together, the seed and Series A rounds therefore bring adsilico’s total equity funding to £13.5M.
With the new funding, adsilico intends to extend its heart model portfolio to cover neurovascular devices, specifically flow diverters for brain aneurysms where trial numbers are very low. The team will grow from 14 to 34 at the University of Leeds’ Nexus innovation hub.
“Patients miss out on the chance of a life-saving device, or face uncertainty about whether it will work for them, because people like them weren’t represented when it was designed and tested. We give medical device manufacturers early insight into the impact of anatomical diversity when it can still shape the design and help avoid costly mistakes or harm to patients,” said chief executive and co-founder Sheena Macpherson.
The gap in the trial data
A JAMA Internal Medicine study of 195 high-risk implant trials published between Jan 2016 and May 2022 found that women comprised a median of 33% of the participants. That figure fell to 29% in cardiovascular trials and 27% in stent trials, but rose to 44% in heart valve trials.
Late-stage mistakes are expensive. The average cost of developing a new complex medical device is around $522 million, including failure, according to a 2022 JAMA Network Open study hosted by the US Department of Health and Human Services.
Regulators are beginning to look at this. An independent UK review has found that some ethnic groups may be receiving sub-optimal care as a result of bias in the design and use of medical devices. In March 2024, the MHRA said it was looking to produce more robust guidance for manufacturers on diversity in testing.
In the US, the FDA’s November 2023 guidance on computational modelling covers physics-based and mechanistic models and, as a trade summary makes clear, doesn’t cover standalone machine learning models. Its September 21 rule on alternatives to animal testing covers drugs and biologics only, not devices. On that reading, the physics step in adsilico’s process is covered by the guidance, and the generative AI step is not.
Who built the virtual patients?
Founded in 2024, the company uses anonymised clinical scans to generate virtual hearts and blood vessels via generative AI. Clinical supervision then subjects them to physics-based simulations of device performance.
The virtual populations can vary by age, sex, ethnicity, body size, and disease. This lets manufacturers assess device performance in a specific population (e.g., women aged 55-75 of South Asian origin) before clinical use. The results can inform device sizing, identify higher-risk patients, and help design trials. The company says it complements lab, animal and clinical studies and may ultimately reduce the need for animal testing.
The founders have very different backgrounds. Macpherson was one of 10 women featured in our Ada Lovelace Day 2024 piece. The other two founders are Dr Nishant Ravikumar and Dr Zeike Taylor, both at the University of Leeds. The technique stems from research funded by the Royal Academy of Engineering and the University of Leeds.
Other players in the space are Virtonomy, which develops virtual patients for device companies and raised over €5M in a Series A in December. Unlearn builds digital twins of trial participants for drug development and raised $50M in a Series C in February 2024. Unlearn is in the drug space; Virtonomy uses actual patient populations and creates fully synthetic populations that are designed to be representative of the real world.
Other investors have also backed the space, with TFN reporting Quibim’s $50M for AI-powered human digital twins in January 2025 and Qureight’s $20M Series B for AI imaging in lung and heart clinical trials in July 2026.
Northern capital and the open question
A £10M round is well in line with the average deal size in the Cambridge life sciences and deep tech space, with most new deals backed by US funds. All three of adsilico’s investors are focused on UK university spinouts. The in silico clinical trials market is projected to be $4.35 billion in 2026 and $12.60 billion by 2040.
“adsilico is a prime example of the ground-breaking innovation and deep thinking that is being generated in universities right across the North of the UK. What Sheena and the team have developed has the potential to make a real global impact for medical device manufacturers, the pre-clinical development process and above all else patients. PXN is very pleased to be joining at such a pivotal point in the company’s development, supporting the team as it scales and sets new benchmarks for in silico testing,” said Dr Elizabeth Young, investment manager at PXN.
The virtual hearts can look very similar to the original scans. The bigger question is whether regulators will accept virtual-patient data as proof for real-world use.
