A new study from the University of Surrey shows that a new ‘out-of-the-box’ way to teach AI models of decision-making could offer hope for new cancer therapeutics.
Computer scientists from Surrey have shown that an open – or model-free – reinforcement learning method is capable of stabilizing large datasets (up to 200 nodes) used in AI models. This approach opens up prospects for uncovering ways to stop cancer progression by predicting the response of cancer cells to perturbations including drug therapy.
Dr Sotiris Moschoyiannis, corresponding author of the study from the University of Surrey, said:
“There is a heartbreaking number of aggressive cancers with little or no information about where they come from, let alone how to classify their behavior. This is where machine learning can provide real hope for all of us.
“What we have shown is the ability of the reinforcement learning-based approach to address real, large-scale reasoning networks from the study of metastatic melanoma. The results of this research have succeeded in using the recorded data to not only design new therapies but also make existing ones more precise. The next step will be to use cells living in the same ways.”
Reinforcement learning is a machine learning method that rewards a computer for making the right decision and punishes it for making the wrong ones. Over time, AI learns to make better decisions.
The model-free approach to reinforcement learning is when the AI has no clear direction or representation of its environment. The model-free approach is more powerful as the AI can start learning immediately without the need for a detailed description of its environment.
Professor Francesca Bova of the Department of Oncology at the University of Oxford commented on the research findings:
“This work is a huge step towards allowing diagnosis of perturbations on gene networks that is essential as we move toward targeted therapies. These results are exciting for my lab as we have long been considering a broader range of perturbations to include the cell microenvironment.”