Pathway analysis of genomic data: concepts, methods, and prospects for future development

Abstract

Genome-wide data sets are increasingly being used to identify biological pathways and networks underlying complex diseases. In particular, analyzing genomic data through sets defined by functional pathways offers the potential of greater power for discovery and natural connections to biological mechanisms. With the burgeoning availability of next-generation sequencing, this is an opportune moment to revisit strategies for pathway-based analysis of genomic data. Here, we synthesize relevant concepts and extant methodologies to guide investigators in study design and execution. We also highlight ongoing challenges and proposed solutions. As relevant analytical strategies mature, pathways and networks will be ideally placed to integrate data from diverse -omics sources to harness the extensive, rich information related to disease and treatment mechanisms.

Figure 1

PubMed citations for “pathway analysis”: 2001-present. The use of pathway analysis has grown exponentially in the last 3–5 years. This explosion in use has followed major developments (shown in boxes) in characterizing the human genome and in performing genome-wide studies of complex diseases and traits. Data points represent the total number of references displayed through a PubMed search for “pathway analysis”, using date limits of January 1, 2001 and December 31 of the calendar year denoted on the x-axis.

Figure 2

An Informed Guide to Pathway Analysis. Broadly, there are two approaches to pathway analysis. In candidate pathway analysis, prior knowledge is used to select pathways hypothesized to have a relationship with a phenotype. In contrast, genome-wide pathway analysis is designed to uncover significant pathway-phenotype relationships within a large data set; insight and prior knowledge are then used to interpret the findings. In both approaches, care must be taken in acquiring pathway annotations and in selecting an appropriate analytical test for association. In addition, other methodological issues (red box) guide the choice of approach and impact strategies for confounding factors. Finally, replication of pathway analysis findings in independent data sets is imperative in validating results to extend their impact.

Figure I

A primer on biological pathways and networks. (a) The major types of biological pathways are shown along with a representation of their relationships among each other. Each type of pathway is defined by its essential goal. Molecular pathways have an essential goal of basic biochemical action (biosynthesis, biodegradation, translocation, transformation, activation, or inactivation) on molecules or compounds. Cellular pathways regulate global cellular status, while organ/system pathways execute higher-order physiological functions. (b) Pathways and networks, while complementary sets of biological elements, differ in key respects. Pathways can include directional regulation (shown in red and green) and branching, but are nevertheless vector-driven to an essential outcome. While elements in pathways are typically connected mechanistically, network elements are connected through shared relationships that may not indicate an action. As such, networks are not vector-driven from a starting point to an essential outcome. Networks can be divided into subnetworks (shown in blue) exhibiting all elements connected to a central node (“A” in this example) or into modules (shown in purple) that exhibit a high density of connections.