New research from the Institute of Cancer Research London (ICR) and the University of Edinburgh has found that a new artificial intelligence-based technique can predict how a cancer tumour evolves and how the disease will progress in a specific patient. Until recently, the mutations which occur in DNA which lead to the progression of cancer have been described as random, but what if this is not entirely true?
Researchers have found that there are recurrent patterns of genomic changes in certain types of tumours which reflect evolutionary processes. These can be used to anticipate the way in which cancer progresses, essentially forecasting the path of the tumour growth. The study used a technique called “Repeated Evolution of Cancer” also known as “Revolver” to find patterns of change in the genetics of tumour cells, which were previously hidden in large, complex data sets.
Until recently, the mutations which occur in DNA which lead to the progression of cancer have been described as random, but what if this is not entirely true?
Scientists developed a new machine-learning method based on “transfer learning” to overcome the challenge of the apparent randomness of tumour changes, filtering out the useful information. They applied this to 768 tumour samples from 178 lung, breast, kidney and colorectal cancer patients from other studies. The technique successfully categorised the patients according to the changes in their tumours and what this meant for their disease progression. It was found that this process works as a “prognosis indicator” allowing the prediction of the clinical changes in the patient, and their survival outcome. In simple terms, it uses information about the progression of one tumour to predict the progression of another.
For example, it was found that a breast tumour containing mutations in the p53 tumour suppressor gene, followed by mutations in chromosome 8 survived for a shorter amount of time than those with other similar genetic changes.
This process works as a “prognosis indicator” allowing the prediction of the clinical changes in the patient, and their survival outcome
With cancer causing one in six of all global deaths, being just one step ahead is a huge leap for modern medicine. The recent rise in drug-resistant cancer tumours has put a strain on current cancer treatment, but this new AI-based method could allow us to foresee the onset of resistance and plan personalised treatment.
Dr Andrea Sottoriva, leader of the study and team leader in evolutionary genomics and modelling at the ICR described this new method as a “powerful artificial intelligence tool” which removes one of cancer’s “trump cards”, the fact that its evolution appears random. Therefore, we can intervene at an earlier stage to improve treatment outcomes. ICR Chief Executive Professor Paul Workman describes the ability of tumours to undergo random genetic change as the “biggest challenge” when finding effective treatment, but being one step ahead of these changes could mean doctors are able to account for drug resistance before it occurs, stopping cancer in its tracks.
The recent rise in drug-resistant cancer tumours has put a strain on current cancer treatment, but this new AI-based method could allow us to foresee the onset of resistance and plan personalised treatment
With the trend of personalised medicine on the rise, treatments which are tailor-made for each individual patient could be the next big break-through in the battle against cancer. Victory may still be a long way off, but developments like this are a key tool for scientists to gain the upper hand in this fight.