Predicting Breast Cancer Survival Rate Based on Genetic Data: A Machine Learning Approach

Breast cancer is the predominant form of cancer affecting women worldwide, and there is a relatively high success rate (70–80%) in treating individuals diagnosed with early-stage, non-metastatic breast cancer. Unfortunately, advanced breast cancer that has spread to distant organs continues to pose a significant challenge and is associated with a lower survival rate, despite the existing treatment modalities. However, since its molecular characteristics were thoroughly identified, There has been a profound transformation in the perception of breast cancer and now encompasses immunohistochemical markers, genomic markers, and immunomarkers, such as TIL and PD-L1. These markers include ER, PR, HER2, and proliferation marker protein Ki-67. This cancer can now be classified into variants using markers thanks to improvements in health care technology, which also make it possible to better detect and characterize the disease. These markers can then be used to predict treatment outcomes, risk of distant recurrence, and prognosis using an assortment of molecular and clinical variables. In this paper, we have tried building a machine learning model that can predict the rate of survival of a patient based on important genetic and clinical markers information. We analyzed the patients’ records and have discovered interesting patterns amongst the patients. Upon further cleaning of the model and working with only clinical data, the best accuracy came with a Logistic Regression mode having 82.5% accuracy. These results can provide a deeper insight into the severity of the disease and can indicate the kind of treatment or therapy that is required or should be recommended to the patient, as the clinical data of patients is constantly measured. Not only such a model can be incorporated in the current treatment of patients, but can also be used to constantly track patient’s progress. However, it is necessary to conduct more validation studies and clinical trials to evaluate the suggested model’s resilience and generalizability in a larger patient population.