Personalised assistance for breast cancer treatment from high dimensional gene expression using hubness aware approaches

Abstract

Breast cancer is the most prevalent type of cancer found in newlineIndian women followed by gynaecological cancers. Breast cancer occurs due newlineto the tumor growth in the breast cells of human body. The risk of death due newlineto breast cancer can be reduced to a greater extent if breast cancer can be newlinedetected during the early stages of cancer and by providing appropriate newlinetreatment. Irrepressible growth of cells occurs due to the mutations that occur newlinein genes controlling the cell cycle. Genes contains the coding to develop newlineprotein required for the functioning of cell from cell growth, repair and newlinecontrol. Some of the gene codes may get altered due to genetic factors or newlineexposure to carcinogens, leading to mutations. Gene expression is the process newlinein which the genes express itself into protein. A mutated gene may lead to the newlineexpression of an abnormal protein leading to various diseases such as cancer. newlineSince gene expression has a high impact to cancer growth, various studies are newlinegoing on to study the relation between cancer phenotypes to gene expression. newlineHence the treatment has to be tailored according to the gene constitution of newlinethe individual. To provide assistance for personalized therapy, the proposed newlineresearch work focuses on developing models for breast cancer detection, newlinebreast cancer subtype prediction and drug sensitivity prediction. newlineSince the gene expression data generated for each patient is newlineextremely high dimensional, learning algorithms would face challenges such newlineas curse of dimensionality, sparsity and presence of outliers. In addition, the newlinegene expression data has large number of features and small number of newlinesamples. Hubness property in high dimensional data is a factor that has been newlineobserved in high dimensional data and could be as a measure in learning newlinealgorithms efficiently. newline