Direnc Pekaslan MSc Dissertation 2014/15

Machine Learning System to recognise cancer from a gene-expression profile

Supervised by D.Walker

Abstract

Cancer is the leading cause of death by disease in the world. In 2012, there were 32.6 million people (within five years of diagnosis) who were suffering from cancerous diseases and 8.2 million of these resulted in death (www1, 2015; www2, 2015). Due to the unique response of each patient to treatment, clinicians need accurate information of diagnosis and prognosis in order to be able to tailor treatment successfully. The purpose of this project is to develop an accurate computational tool which can predict information such as the stage, metastasis capability and/or typology of cancer from a publicly available gene-expression profile based on machine learning techniques.   In this report, relevant literatures that have used a multilayer neural network in gene expression datasets to classify and predict survivability and identify biomarkers of cancer are investigated . A summary of the main findings suggests that a multilayer neural network is capable of accurate classification and prediction in cancer gene expression profiles.