Microarray gene expression profile data mining model for clinical cancer research

TL;DR: A newly designed data mining model is proposed, fashioned from a computer science point-of-view, to store microarray experimental data in a systematical organization, and to provide an efficient way for researchers to mine the database and populate it in a reasonable manner as their research progresses.

Abstract: The DNA microarray is the latest breakthrough in molecular biology, which provides researchers with an approach to monitor genome-wide expression systematically. Its application in cancer study has proved to be successful in elucidating the pathological mechanism, with the potential of altering clinical practice through individualized cancer care and ultimately of contributing to the battle against cancer. However, the current hurdle and challenge is how to make use of the tremendous amount and ever-growing microarray experimental data to precisely explain the cancer mechanism and to better predict the cancer development in the early stage. This topic has been realized by traditional biologists and presented to a new group of scientists from Biology, Statistics and Computer Science. We propose a newly designed data mining model, fashioned from a computer science point-of-view, to store microarray experimental data in a systematical organization, and to provide an efficient way for researchers to mine the database and populate it in a reasonable manner as their research progresses. The model in our design addresses the interpretation of the meaning of the microarray gene expression profile data in cancer research in the context of the biological pathway, with focus on the elucidation of key pathways in cancer development, thus providing a bridge between clinical cancer research and microarray gene expression raw data. An object-relational database schema is proposed, which includes six subsystems: array, cancer, drug, gene, image and pathway. The relationship between the gene expression profiles under different experiment conditions and biological processes can be drawn from this database. This newly designed data mining model provides an efficient way to translate the large collection of existing profiles so as to be a handy reference for clinicians who face cancer early detection, clinical diagnosis and treatment decisions; it offers a new paradigm as a patient education tool for better patient care and health advisory against human disease; it also provides molecular biologists with an alternative and feasible route to interpret genetic experimental data, which may ultimately lead to a more complete understanding of a complex human disease-cancer.