UCL Scientists Develop Digital Twins to Predict Cancer Treatment Responses

Researchers at University College London and the University of Cambridge developed a symbolic artificial intelligence model that creates digital twins of patient tumors, accurately predicting treatment responses for triple-negative breast cancer in 74% of cases during a 31-patient test.

Hard-to-treat cancers often leave clinicians searching for reliable ways to match patients with therapies before standard interventions fail. Researchers turned to computational modeling to simulate how individual tumors react to specific drugs. The findings, published in Cell Reports Medicine, demonstrate that virtual replicas can map out complex biological responses using detailed patient genetic data according to a UCL announcement.

Simulating Trial Data

The PARTNER trial ran between 2016 and 2024. The research team built their computer-generated tumors using genetic information from participants in the Platinum and PolyAdenosine 5′-Diphosphoribose Polymerization Inhibitor for Research and Treatment of Neo-Adjuvant Early and Relapse-Risk Breast Cancer clinical trial.

Scientists at the University of Cambridge and University College London designed a software system that generates virtual copies of patient tumors to evaluate which medications can successfully eradicate the illness. Triple-Negative Breast Cancer (TNBC) makes up about 15% of all breast cancer cases. It multiplies at a high speed, undergoes frequent mutations, and lacks the trio of common markers utilized by conventional hormone treatments to combat cancer cells.

UCL Scientists Develop Digital Twins to Predict Cancer Treatment Responses — Photo: WBAY

Triple-negative breast cancer accounts for roughly 15% of all breast cancer cases. The disease grows quickly, mutates rapidly, and lacks the three standard receptors—for oestrogen, progesterone, and the HER2 protein—that traditional hormone therapies and targeted drugs rely on. Within the simulation, researchers tested regimens mirroring the trial: chemotherapy alone, chemotherapy combined with a drug named Olaparib, and a no-treatment scenario as detailed in study reports.

Comparing AI Models

Many clinicians remain skeptical of standard artificial intelligence platforms because large language models operate as opaque black boxes where the inner workings remain hidden. The UCL team utilized symbolic AI, a model driven by logical reasoning rather than probabilistic guessing. Numerous medical professionals harbor doubts regarding traditional artificial intelligence tools because large language models function as a “black box,” obscuring the way the program arrives at a conclusion.

The investigators deployed “symbolic AI” instead, a framework grounded in sequential biological logic rather than mere estimation, enabling physicians to trace every phase of its deduction.

We have shown the broad and powerful range of insights that personalised virtual tumours could bring to cancer medicine through our symbolic AI-based computer model.

Professor Jasmin Fisher, lead author at the UCL Cancer Institute

Because the system relies on step-by-step biological logic, doctors can inspect and trace every step of the program’s reasoning noted the researchers. Fisher emphasized that the software is designed to serve as an aid for existing clinical practice rather than an automated replacement. Fisher added, “We see our computer model as an aid for existing cancer clinical practice, not an alternative to it: a way of generating new ideas particularly in relation to proteins or genes that could be targeted in ways that, as it were, restore a patient’s sensitivity to existing drugs and make those treatments effective.”

Futuristic techno design on background of supercomputer data center timofeev vladimir — Photo: News Medical

Targeting Non-Responders

When the simulation results were compared against actual clinical trial outcomes, the computer program correctly predicted whether treatment would work in 74% of the cases. The digital twins offered actionable insights for patients whose real-world tumors failed to shrink under standard care, and suggested ways to help patients whose tumors had not responded well to treatment.

The simulations forecast that incorporating a precision therapy would shrink the tumor to such an extent that no active cancer would be left behind in the lymph nodes or breast tissue. This was found in the case of 10 such participants.