Using AI to predict drug resistance mutations in breast cancer

Meet the scientist

Beyond being a breast pathologist, Dr Anne Vincent-Salomon is a mom of 3 and a passionate hiker. She loves being outdoors, hiking in the South of France or in Corsica, and in the streets of Paris. 

The science

Around 70% of breast cancers are diagnosed as ER-positive, which means that the cancer cells produce too much of a protein called the oestrogen receptor and become dependent on the hormone oestrogen to grow. Therapies that target this dependence on oestrogen, such as aromatase inhibitors, are a common treatment for this type of breast cancer but some people can develop resistant to treatment over time. Research has shown that one way this resistance develops is due to mutations to a gene called ESR1, which seem to emerge in response to treatment with aromatase inhibitors. 

Dr Anne Vincent Salomon and her team are studying whether there are characteristics of breast tumours that are visible prior to treatment, which could be used to predict who will develop mutations to the ESR1 gene, and even predict who will develop resistance to treatment. The team will use images captured from thin slices of tumours donated from breast cancer patients to train a deep learning artificial intelligence tool. They hope that this tool will identify predictive characteristics of breast tumours that will be able to improve how patients are treated in the future.