Statistical Aspects of ChIP-Seq Analysis

Abstract

Introduction: The Purpose of the ChIP-seq Experiment Every cell is host to a diverse ecosystem of proteins, each protein having its own functional properties. Cell behaviors and processes, such as growth, are dependent on the levels of these various proteins. Through microarrays, and subsequently high-throughput sequencing, we have become adept at quantifying the expression levels of genes, in various cell types under different conditions. These expression levels provide us with a proxy, albeit an imperfect one, for the levels of protein production in a cell. Indeed, estimates of the correlation between mRNA levels and protein levels are typically very variable [Gry et al., 2009]. We can infer links between cause and effect by perturbing a cell in some way and measuring which genes' mRNA levels change in response. However, this information is, by itself, unsatisfying – specifically, we seek to understand the regulatory mechanisms underlying these links. This has clinical significance – ultimately, we may be able to manipulate these mechanisms ourselves, potentially leading to novel therapies. Consider the following example: in breast cancer, malignant tumors typically exhibit an invasive, proliferative behavior characterized by abnormal growth (Weinberg, 2007). A critical factor in establishing this behavior is the estrogen receptor (ER). In particular, ER is known to encourage mitotic cell division. This function is, in itself, not dangerous – cells in healthy tissue commonly divide to replace lost cells. However, ER is known to be a key factor in establishing cancer, and approximately 70% of breast tumors are labeled as ER-positive – that is, their cells have ER content above a certain threshold (Tannock and Hill, 1998; Mohibi et al., 2011).