A data integration methodology for systems biology: Experimental verification

Abstract

The integration of data from multiple global assays is essential to understanding dynamic spatiotemporal interactions within cells. In a companion paper, we reported a data integration methodology, designated Pointillist, that can handle multiple data types from technologies with different noise characteristics. Here we demonstrate its application to the integration of 18 data sets relating to galactose utilization in yeast. These data include global changes in mRNA and protein abundance, genome-wide protein-DNA interaction data, database information, and computational predictions of protein-DNA and protein-protein interactions. We divided the integration task to determine three network components: key system elements (genes and proteins), protein-protein interactions, and protein-DNA interactions. Results indicate that the reconstructed network efficiently focuses on and recapitulates the known biology of galactose utilization. It also provided new insights, some of which were verified experimentally. The methodology described here, addresses a critical need across all domains of molecular and cell biology, to effectively integrate large and disparate data sets.metabolism ͉ yeast ͉ molecular network model ͉ galactose S ystems biology aims to understand the dynamic behavior of molecular networks in the context of the global cell, organ and organism state by exploiting (i) high-throughput interrogation technologies; (ii) increasingly comprehensive databases of biomolecules and their interactions; and (iii) computational predictions of molecular function and interaction ( Fig. 1). Use of each of these sources of information has its own drawbacks (1). For example, many current global assays of mRNA and protein abundance͞state are systematically biased toward more abundant species and measure only the average content of many thousands of cells. Global assays are also inherently noisy and include significant numbers of false positives and false negatives. Databases tend to combine data from different cell types, different strains of an organism, and different experimental conditions. Moreover, well studied molecules and pathways are systematically overrepresented in databases. As a result, integration of database information with a particular set of experimental data can introduce systematic biases into the modelbuilding process. Likewise, computer predictions tend to be more accurate for members of well characterized molecular families. Therefore, there is a pressing need for data integration methodologies that effectively address both random noise and systematic bias in data.In a companion paper (2), we present a data integration methodology and its software implementation to address these challenges. To present the methodology clearly, we used only simulated data in that paper. Here, we present the application of our methodology (named Pointillist) to 18 types of biological data, which we integrate to arrive at a detailed and comprehensive picture of galactose utilization in yeast. The data we integrate include in...

A data integration methodology for systems biology

“…The applicability of our methodology to different types and sizes of data and to different numbers of data sets is demonstrated by application to five different types of data integration in a companion paper (10). Although we focused here on presenting our methodology from the perspective of maximizing statistical power, it can also be applied to scenarios for which the different types of data being integrated have systematic differences between them, for example, combining mRNA and protein abundance measurements or in vivo and in vitro measurements.…”

“…Although we focused here on presenting our methodology from the perspective of maximizing statistical power, it can also be applied to scenarios for which the different types of data being integrated have systematic differences between them, for example, combining mRNA and protein abundance measurements or in vivo and in vitro measurements. Examples of this type of integration are given in the companion paper (10). Data integration can never rule out inclusion of some false positives or loss of some true positives.…”

Construction of gene/protein interaction networks for primary myelofibrosis and KEGG pathway-enrichment analysis of molecular compounds

“…There are three main categories to construct biomolecular regulatory networks: gene regulatory networks through a mathematical model; networks through literature mining, and integrating multiple data (Friedman et al, 2000;Hwang et al, 2005;Mohamed-Hussein and Harun, 2009). Building a network through literature mining means using bioinformatics, computational biology, and other tools of computer science to analyze the data in the literature, and build biomolecular regulatory networks using the relationships between gene/protein interactions of the existing literature.…”

Systems Biology: The synergistic interplay between biology and mathematics

“…Hwang et al (2005a) have also recognized the issues in data integration, when the different data types vary in size, confidence, and network coverage and have developed a general algorithm based on advanced statistical techniques to better interpret the data. These authors used the developed approach to analyze 18 data sets including mRNA, protein levels, protein-DNA interaction data, and protein-protein interaction data related to galactose utilization in S. cerevisiae (Hwang et al, 2005b) and identified 69 genes that were perturbed significantly in the data sets. Additionally, the analysis suggested that fructose metabolism would be down regulated in the presence of galactose via the downregulation of a hexose transporter, a hypothesis that was experimentally verified through the measurement of corresponding protein levels.…”