A locus-based paradigm for generating systems biological inferences from large scale functional genomics datasets

Genomics data is growing at a exponential rate. The ability to integrate new results with existing knowledge about genomic biology is rapidly becoming the limiting factor as there no universal language with which to describe genomic functional elements. To integrate and compare new and existing genomic data, we define our basic functional unit of a genome to be a locus---a set of positional coordinates along any genome with an arbitrary amount of functional annotations attached. The locus concept enables addressing genomic elements and annotations at any level of granularity from entire swaths of chromosomes to single base-positions. We define a locus-based framework to compare a given set of genomic elements to any of existing genomic annotations. We use this to build a tool to find genomic annotations significantly and frequently overlaping with a set. We also use this to build a tool to infer functional interactions from locus intersections and show how the inference of regulatory interactions from genomics data and the analysis of the topological properties of genomic networks can provide useful biological insights.;We demonstrate the importance of the locus based paradigm in the emergent field of ribonomic profiling. We show how locus based comparisons reveal novel overlaps between the binding sites of microRNAs and the HSL-mRBP and explore other potential coordinate regulatory sites. We use locus-based intersections to compare RIP-Chip derived targets of the HuR mRNA binding protein across cell-lines and platforms to each other, to known HuR targets in the literature, and to AREs thought to be associated with HuR targets. We reveal a startling lack of overlap between targets from matched samples profiled on different platforms and across the target sets in general. This shows the need for further refinement on RIP-Chip technology and highlights the efficacy of the locus-based approach. Finally, we profile well-characterized sets of RBP-binding sites against sets of regulatory, functional, and disease annotations, and find that we can not only recapitulate the known functional properties of these sites but also find relatively unknown aspects of these sets that are supported by literature.