Modular architecture in biological networks

Abstract

In the past decade, biology has been revolutionized by an explosion in the availability of data. Translating this new wealth of information into meaningful biological insights and clinical breakthroughs will require a complete overhaul both in the questions being asked, and the methodologies used to answer them. One of the largest challenges in organizing and understanding the data coming from genome sequencing, microarray experiments, and other high-throughput measurements, will be the ability to find large-scale structure in biological systems. Ideally, this would lead to a simplified representation, wherein the thousands of genes in an organism can be viewed as a much smaller number of dynamic modules working in concert to accomplish cellular functions. Toward demonstrating the importance of higher-level, modular structure in biological systems, we have performed the following analyses: 1. Using computational techniques and pre-existing protein-protein interaction (PPI) data, we have developed general tools to find and validate modular structure. We have applied these approaches to the PPI networks of yeast, fly, worm, and human.

(cont.) 2. Utilizing a modular scaffold, we have generated predictions that attempt to explain existing system-wide experiments as well as predict the function of otherwise uncharacterized proteins. 3. Following the example of comparative genomics, we have aligned biological networks at the modular level to elucidate principles of how modules evolve. We show that conserved modular structure can further aid in functional annotation across the proteome. In addition to the detection and use of modular structure for computational analyses, experimental techniques must be adapted to support top-down strategies, and the targeting of entire modules with combinations of small-molecules. With this in mind, we have designed experimental strategies to find sets of small-molecules capable of perturbing fimctional modules through a variety of distinct, but related, mechanisms. As a first test, we have looked for classes of small-molecules targeting growth signaling through the phosphatidyl-inositol-3-kinase (PI3K) pathway. This provides a platform for developing new screening techniques in the setting of biology relevant to diabetes and cancer. In combination, these investigations provide an extensible computational approach to finding and utilizing modular structure in biological networks, and experimental approaches to bring them toward clinical endpoints.