Estimation of Resemblance and Risk Level of a Breast Cancer Patient by Prognostic Variables Using Microarray Gene Expression Data

Breast cancer is a common type of cancer affecting women worldwide. Continuous efforts are being made for the identification of significant genes for prognosis of breast cancer. A microarray gene expression dataset contains tens of thousands of genes. Identification of smaller subset of disease-causing genes from a large gene expression dataset is a challenging task for the researchers. Here, a variant of Cox proportional hazard regression model, namely 1d-DDg method, has been applied to select the predictive genes for breast cancer patients. Here, in this paper, instead of using the Euclidian distance, the Manhattan distance has been used to estimate the risk level of a query (newly admitted) patient by matching similarity with the reference (existing) patients. The time complexity of Manhattan distance is better compared to Euclidian distance. The post-surgery disease recurrence and the level of risk of a breast cancer patient can be reliably predicted using personified prognosis.