Healthcare (Commonwealth Union) – A computer tool developed by scientists in the UK makes a virtual copy of a patient’s tumour and then tests different treatments on it. They hope it will help doctors pick better care for people with some of the most difficult cancers.
What the researchers did
A team at University College London (UCL) has developed a program that uses a patient’s detailed genetic fingerprint to predict which treatments could kill their cancer, particularly when standard treatments have failed, working with colleagues at the University of Cambridge. The work was published in the journal Cell Reports Medicine .
The team investigated triple-negative breast cancer, or TNBC. It is a fast-growing, aggressive form of the disease that makes up approximately 15% of all breast cancer cases. It is difficult to treat as it grows rapidly and has fewer targets for treatment.
How Do Digital Twins Work?
To test their idea, the researchers created computer-generated tumours, called “digital twins.” These were replicas of the actual tumours from 31 patients who participated in a clinical trial called PARTNER.
Then the team repeated the trial on-screen. They ran simulations of treatment with chemotherapy alone and with chemotherapy plus a drug called Olaparib, known to work against TNBC in some cases, on each digital twin. They also did a simulation with no treatment for comparison.
The results were promising. The researchers then compared the simulations to the actual trial results and the program accurately predicted whether treatment would work 74% of the time.
Finding help for non-responding patients
The digital twins also identified strategies to help patients whose cancers didn’t respond well. In 12 of the 13 patients who did not fully respond, the models identified at least one additional targeted treatment that would have been able to improve the outcome. For 10 of these patients, the models predicted that the additional treatment could have generated a sufficiently strong response that there likely would have been no invasive cancer remaining in the breast and lymph nodes afterward.
Since the rapid evaluation of technology, the ability to run simulations and forecast the possible outcomes have become more feasible.
Researchers of the study stressed that the computer forecasting, was not proof that the treatments would be successful. This would require tests prior to it becoming available for patients.
Most of us are aware of AI tools that are difficult to track in their inner workings. The scientists indicated that some medical professionals have been sceptical due to the fact that AI is like a “black box” in that it is not precise into the methods it draws conclusions.
The UCL researchers took a different route. They applied “symbolic AI”, which is based on logical reasoning instead of prediction, and whose operation is clearly transparent. So simply put, they are able to see the steps the program carried out, and they have the ability to challenge them. Professor Jasmin Fisher at the UCL Cancer Institute, who is the lead author on the study says ““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.”
Professor Fisher said the tool was designed to assist doctors, not replace them. She said it was a tool to help with existing cancer care and a way to generate new ideas, especially about proteins or genes that could be targeted in new ways to help a patient’s cancer respond to treatment.
She also admitted that the program is not always going to be right. A doctor might look at a suggested treatment very carefully and decide not to use it, for example if the drug is too toxic for that particular patient. So simply put, the medical team still has the last word.
With just 31 patients enrolled in the study, bigger trials will be needed to confirm the findings. Before the tool can be used in hospitals, the researchers will also need to show it works across different patients and different types of cancer.
But early results seem promising. If the approach works, doctors may eventually be able to test treatments on a patient’s digital twin before deciding what to give the patient. It could save time, avoid unhelpful treatments and offer new hope to people whose cancer has resisted standard care.
