The Human Genome Project recently announced the sequencing of the human genome with enormous potential implications such as customized medicine. New analytic methods and computational technologies are needed to fulfill this promise. This study focuses on the development and integration of signal processing approaches for aiding the diagnosis and treatment of diseases such as cancer. The massive amount of data generated by high-throughput microarray technology provides a gateway to the discovery of key genes and gene combinations that explain specific diseases on a mechanistic level, to the classification of diseases on a molecular level, and to the development of optimal therapeutic strategies for various diseases. This study is concerned with (i) developing signal processing based solutions to individual sub-problems that arise in functional genomics; and (ii) ensuring that these solutions can be pieced together to provide solutions to the original overall problems of disease diagnosis and therapy.
Microarray-based genomic signal processing can be broadly broken into five categories: signal extraction, clustering, classification of phenotypes (diagnosis), modeling genetic regulatory networks, and developing strategies for regulatory intervention (therapy). There tends to be a natural hierarchy among the five categories which are, therefore, studied from a systems perspective to ascertain the manner in which processing, analysis, and estimation on one level propagates through the system to affect downstream analyses, and how the subordinate issues within a category interact among each other. These questions are addressed within a unified theoretical framework, their solutions are applied to real biological problems, and a large body of integrated software to facilitate further investigation and application is developed.
