Novel biology driven bioinformatics methods in cancer genomics

NOVEL BIOLOGY-DRIVEN BIOINFORMATICS METHODS IN CANCER GENOMICS YU KUN (M.Science, NUS) A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF PHYSIOLOGY YONG LOO LIN SCHOOL OF MEDICINE NATIONAL UNIVERSITY OF SINGAPORE 2008 Acknowledgements I would like to express my sincere gratitude and appreciation to my mentor Patrick Tan, for his enthusiastic encouragement, excellent guidance and kind support throughout the duration of this research. Probably more importantly, as a great mentor, Patrick ignites my passion for the research and gives me a very joyful science trip. I am indebted to Kumar Ganesan and Yansong Zhu, who did the siRNA knockdown experiments. I also thank Amit Aggarwal. With Amit and Kumar, I spent innumerable hours in discussing the issues of cancer genomics and such related topics have been a major influence in my vision of the entire field. I would like to extend my deep appreciation towards Jeanie Wu, who did the microarray profiling of tumor samples. Without her great effort, I simply cannot start my project. I would like to express my deep appreciation on the help from Lay Keng Tan (Angie) and Benita K T Tan. With their great effort, we successfully finished the clinical validation of a series of breast cancer gene signatures. I am grateful to Dr. Tan Puay Hoon, Dr. Hong Ga Sze, Dr. Wee Siwe Bok, Dr. Chow Yin Wong, Dr. London Lucien Ooi, Dr. Thirugananam Agasthian, Dr. Wai Keong Wong and Dr. Khee Chee Soo for sample collection. II I would like to thank our collaboration lab from Department of Physiology, National University of Singapore: Mirtha Laban, Hongmin Li, Carol Ho Wing Leung, and Shing Chuan Hooi; for their work on in vivo cancer metastasis model. I would also like to thank our collaboration team from Genome Institute of Singapore: Xiao Dong Zhao and Chia Lin Wei; for their work on ChIP-PET of MYC. I would like to acknowledge Dr. Lance D Miller from Genome Institution of Singapore. Dr. Lance D Miller gives us lots of valuable suggestions and comments on our work. He also kindly provides breast cancer data set for our validation. The work was partially supported by grants BMRC 01/1/31/19/209 and BMRC 05/1/31/19/423 to PT from the Agency for Science, Technology, and Research (A-star) I wish to acknowledge Department of Physiology, Yong Loo Lin School of Medicine in National University of Singapore and the University itself, for providing various supports from course education to the management of student affair. I acknowledge and appreciate the love and support of my family. Especially, I would like to give my special thanks to my dear wife May whose love enabled me to complete this work, and I dedicate this thesis to her. III Table of Contents Acknowledgements .II Table of Contents IV Summary VII List of Figures . IX List of Tables XI Abbreviations .XII Chapter 1. Introduction – Gene Expression and microarray . 1.1 Gene Expression . 1.2 microarray . 1.3 Methods for microarray Data Analysis . 1.4 Limitations of Conventional microarray data analysis . 1.5 Thesis Organization 11 Chapter 2. microarray Studies of Cancer Genomics . 13 2.1 Cancer . 13 2.2 microarray Application in Cancer Research . 15 2.3 Novel data analysis methods to refine the molecular profiling of cancer . 21 Chapter 3. Tumor Modules in Breast Cancer . 23 3.1 Breast Cancer 23 3.2 Signature Algorithm (SA) . 24 3.3 Tumor Module (TuM) . 26 IV 3.4 Breast Tissues and microarray Profiling . 27 3.5 Identification of Tumor Modules in Breast Cancer 28 3.6 A Novel Module – TuM1 . 31 3.7 TuM1 is A Robust Module . 34 3.8 TuM1 is an Apoptotic Tumor Module and Associated with Low Histologic Grade in an ER-independent Fashion . 35 3.9 Validation of Association between TuM1 Low Histologic Grade in Multiple Independent Test Sets 40 3.10 TuM1 Module is Down-regulated Upon Tamoxifen Treatment In Vitro . 43 3.11 TuM1 Member Gene relaxin Mediates Tamoxifen Response in MCF7 Cell Line . 46 3.12 A Correlation between TuM1 Expression and Clinical Outcome of ER+ Patients 47 3.13 A Prospective Clinical Validation of TuM1 Using a Customized microarray . 50 3.14 Summary . 52 Chapter 4. A Precisely Regulated Gene Expression Cassette Potently Modulates Metastasis and Survival in Multiple Solid Cancers 54 4.1 Precise Regulation and Ultrasensitivity 54 4.2 Gene Expression Analysis to Identify Ultrasensitive Components 56 4.3 A Conserved Cassette of Precisely Regulated Gene Expression in Multiple Solid Tumors 57 4.4 Identification of PGC is NOT due to the Artifact of microarray Data Analysis 61 4.5 Failure of PGC Detection by Conventional microarray Analysis Methods 64 4.6 A Cross-Validation Assay Confirms Specificity and Robustness of the PGC Signature 66 4.7 Independent Validation of the PGC in Diverse Solid Tumors . 68 4.8 PGC Genes are Associated with Multiple Cancer Related Pathways . 71 4.9 Subtle Alterations in PGC Expression Are Associated with Metastatic Capacity of Cancer Cells 73 V 4.10 siRNA-mediated Knockdown of Multiple PGC Genes Enhance Cellular Invasion . 77 4.11 Subtle Alterations in PGC Expression are associated with Survival Outcome of Cancer Patients . 79 4.12 PGC is Independent From Previous Published Gene Signatures 82 4.13 PGC is Possibly Transcriptional Regulated by Myc . 83 4.14 Summary . 84 Chapter 5. Conclusion and Discussion . 86 5.1 Conclusion 86 5.2 Biological Function of Genes Found in TuM1 . 86 5.3 TuM1 Predicts Tamoxifen Response of ER+ patients 88 5.4 Methodology discussion – Biclustering 89 5.5 Integrative Systems Biology – Go beyond Gene Expression . 90 5.6 PGC Genes are Involved in Carcinogenesis . 90 5.7 The Cancer Robustness . 93 5.8 The Normal-Specific PGC 93 Bibliography . 95 Appendix Tables . 108 Appendix Figures 120 VI Summary In the past decade, numerous groups have reported studies using genome-wide gene expression data generated from microarray. One of the blooming fields for microarray application is cancer research. Despite the promising nature of these initial microarray studies, such conventional approaches are associated with certain limitations. Because of these challenges, it is important to develop new methods to mine the inherent richness of information present in genome-wide expression data, in order to further identify novel, robust, and biologicallyrelevant molecular signatures for the purposes of tumor classification and patient stratification. We applied the signature analysis (SA), which was designed to overcome the limitations of conventional clustering approaches, to a set of breast cancer expression profiles and successfully defined multiple ‘tumor modules’ (TuMs), each associated with a distinct biological function. Most significantly, the SA identified a previously-unreported module (TuM1) in a subset of Estrogen Receptor (ER+) tumors containing genes significantly enriched in cell death and apoptosis. The TuM1 module is not discernible by conventional clustering analysis; and proved to be a robust signature by repeated random sampling assays. We further show that tumors expressing the TuM1 module are associated with low histologic grade (P[...]... targeted therapeutics In a summary, microarray studies in cancer enhance our understanding of cancer genomics and shed the light on clinical practice of cancer diagnosis and treatment 2.3 Novel data analysis methods to refine the molecular profiling of cancer microarray data of cancer cells have been extensively analyzed and led to numerous breakthrough findings However, as we discussed in the section 1.4,... permitting complex phenotype of cancer cells to be defined at molecular genetic 14 level DNA microarray has been widely used in cancer research In the section 2.2 we will discuss the applications of DNA microarray in cancer research 2.2 microarray Application in Cancer Research microarray technology greatly propelled the gene expression study In the past decade or so, microarray has been widely used in cancer. .. unknown In gene-wise clustering, the functions of these genes may be inferred by looking at genes with known functions which belong to the same 7 group This type of inference is based on the guilt-by-association principle For example, if Gene X, a gene with unknown function, belongs to the same group as a gene known to be involved in promoting cell growth, we may infer that Gene X is also involved in promoting... the identification of novel tumor module in breast cancer The biological significance and clinical implication of the tumor module is also addressed in the Chapter 3 The main body of this work has been published in the Clinical Cancer Research (19) A paper that is in press reported a prospective clinical validation of the tumor module signature (20) Chapter 4 provides another novel angle to explore... biological and clinical –relevant information in cancers, even from data sets that have received substantial prior analysis Moreover, the novel methods and the expression patterns identified by these methods are generally applicable to other diseases 22 Chapter 3 Tumor Modules in Breast Cancer 3.1 Breast Cancer Breast cancer is a significant cause of worldwide morbidity and mortality in females (37)... certain limitations associated with these conventional analysis methods Unsupervised learning algorithms, such as Hierarchical clustering, typically cluster genes based on their global expression patterns across all samples (eg, tumors); while in reality 23 certain genes may only show strong regulation in a certain set of tumors, and weak to minimal regulation in others (14, 15) Standard clustering methods. .. are certain limitations of the conventional microarray data analysis methods To further extract the novel knowledge embedded deeply in the richness of microarray gene expression data, we developed and applied two novel data analysis approaches on the microarray data in the context of cancer study We will demonstrate, in the next two chapters, that the novel methods successfully 21 revealed novel biological... subtypes of cancer with not only heterogeneous molecular patterns, but also distinct characteristics to clinical outcome The subtypes have been repeatedly observed in an independent data sets (24) Our group has also reported the conservation of breast cancer subtypes in an AsianChinese patient population (25) The subtypes have also been demonstrated to be presented in pre-invasive stage of carcinogenesis... morbidity and mortality in females (37) A major challenge in the diagnosis and treatment of breast cancer is its heterogeneity, as individual breast tumors can exhibit tremendous variations in clinical presentation, disease aggressiveness, and treatment response (38) For example, one important factor in clinical breast cancer classification include determining the estrogen receptor (ER), as ER is a prognostic... (both mRNA and protein level) are also common in cancer development New aspects of the genetics (or epigenetic) of cancer pathogenesis, such as DNA methylation, and microRNAs are increasingly being recognized as important 13 Figure 2.1 The hall markers of cancer This figure is adopted from Hanahan, D and Weinberg, RA (22) Six fundamental cellular properties can be altered during carcinogenesis, to give . Cancer Genomics 13 2.1 Cancer 13 2.2 microarray Application in Cancer Research 15 2.3 Novel data analysis methods to refine the molecular profiling of cancer 21 Chapter 3. Tumor Modules in. NOVEL BIOLOGY-DRIVEN BIOINFORMATICS METHODS IN CANCER GENOMICS YU KUN (M.Science, NUS) . associated with clinical outcome in multiple cohorts. These findings support the existence of a common set of precisely-controlled genes in solid tumors. By identifying novel clinically-relevant