Rank based statistics in analyzing high-throughput genomic data

Acknowledgments I would like to thank my advisors Dr. Zohar Yakhini from Agilent Labs and the Technion and Prof. Benny Chor from Tel Aviv University for their guidance and support. I would especially like to thank Zohar for giving me the opportunity to work in this exciting field of computational biology. Specifically, for his dedicated attention, his inexhaustible patience, for the captivating numerous discussions on math, computer science, biology and their integration and for teaching me that making science is not limited only to the borders of the office/lab. Special thanks goes to my lab member – Israel Steinfeld, for his friendship, for being an important part of all the studies described in this thesis and for numerous fruitful discussions. During my studies I found myself battling cancer, not only scientifically. I would like to thank my colleagues, advisors, Agilent Technologies Israel, friends and family for being there for me when I needed them the most. Last, but not least, I would like to thank my parents Ruth and Gil Navon for raising me to be curious and for showing me the passion for science and learning. Abstract High throughput methods (such as microarrays) have revolutionized experimental biology. New statistical methods that facilitate analysis of the vast amounts of data generated by these experiments became a critical part of general study workflow. Gaps and variants that address specific situations are the subject of ongoing research and development efforts. Such methods were, indeed, developed and optimized to address various aspects of high-throughput genomic data analysis. This thesis focuses on specific aspects of expression data analysis. Chapter ‎ 2, describes GOrilla-a GO enrichment tool that provides statistically sound methods for identifying GO terms enriched at the top of a ranked list of genes. Chapter ‎ 3 addresses semi-supervised class discovery – finding meaningful partitions in the data based on genomic and phenotypic data. The third method, presented in Chapter ‎ 4 – RCoS, is a rank consistency score, which is useful in analyzing matched data. The statistical methods presented in the thesis are not limited to the analysis of gene expression, but the examples used are based on gene (and miRNA) expression. The methods presented herein are currently in use by a growing number of bioinformaticians and data analysis professionals. For example, GOrilla had over 5,000 visits in the first year following the publication of the paper.