Development of Improved Cancer Classification Method by Integrating Metadata in Microarray Data Analysis

Abstract

Microarray experiments provide a high throughput to measure expressions of thousands of genes simultaneously. A systematic and computational analysis of this vast amount of data provides understanding and insight into many aspects of biological processes like Single Nucleotide Polymorphism (SNP), genetic disorders, cancer identification. Expression analysis of DNA (Deoxyribonucleic Acid) of microarray experiments is far from straightforward from statistical point of view. Dimensionality problem of microarray data, identifying significant and informative genes or DNA sequences are the prime challenges for cancer classification. Therefore, the aim of the thesis in cancer classification is to integrate metadata for optimal subset of genes that are useful for expert and embedded system design for different types of cancer classification. In this thesis, different filtering and classification algorithms will be compared on different set of data to conclude which will be efficient and effective method for cancer classification from microarray data. For the improvement of cancer classification performance and biological validation of optimal subset of genes, metadata ranking will be evaluated and integrated for cancer classification. The method will achieve better performance based on gene-independent covariance, trustworthy gene feature ranking and metadata ranking factors. The performance of the proposed method will be evaluated by different filtering to overcome the dimensionality problem, minimum number of genes used in classification techniques on publicly available benchmark dataset of ALL, brain, breast, kidney, lung, prostate for intracancer and inter-cancer classification.