It is estimated that for 2020 in the US the expected number of new cases of breast cancer in female patients is about 276,000 (30% of all new tumor cases in female patients) and the expected number of deaths caused by breast cancer in female patients is about 42,000 (15% of all deaths due to tumors in female patients), thus making breast cancer the first type of cancer for the number of new cases, and the second type of cancer as cause of death in female patients. Similar rankings are observed in Europe and China.
For a patient with breast cancer who has undergone surgery, it is essential to decide on a post-operative treatment that prevents the recurrence of the tumor disease and the formation of metastases. Bioinformatics intervenes here, alongside the so-called precision medicine, precisely to help oncologists calculate the real effectiveness of the proposed treatments and the patient’s survival prospects.
A study signed by Marco Pellegrini, IIT-CNR Research Director, has recently been published on Scientific Reports, a journal of the Nature group, and can represent an important milestone in this sector. The article, entitled “Accurate prediction of breast cancer survival through coherent voting networks with gene expression profiling”, describes how Pellegrini identified a new classification method supervised by machine learning. Pellegrini applied it to survival prediction at 5 years for patients with surgically removed breast cancer, using gene expression levels measured from tumor samples and obviously taking into account post-operation therapies. More than 2000 data from a list of cancer patients were used to train, validate and test artificial intelligence: the final result indicates a predictive capacity superior to that of the other methods currently in use.
“Our methodology followed two directions”, explains Marco Pellegrini. “On the one hand we drew on genetic sequencing and biomarkers of excised tissue samples; on the other hand we inserted and analyzed these data in a “predictor”, an artificial intelligence tool based on a new algorithm. This made it possible to achieve a prediction accuracy of 80% – 90%”.
This new methodology identified by Pellegrini can provide an important contribution to clinical decisions on breast cancer therapy and give doctors an additional tool to personalize therapies and increase the chances of survival of their patients. For this reason, the discovery was filed for a patent application in Italy, the United States and the European Community.