Discovering molecular components and their functionality is key to the development of hypotheses concerning the organization and regulation of metabolic networks. The iterative experimental testing of such hypotheses is the trajectory that can ultimately enable accurate computational modelling and prediction of metabolic outcomes. This information can be particularly important for understanding the biology of natural products, whose metabolism itself is often only poorly defined. Here, we describe factors that must be in place to optimize the use of metabolomics in predictive biology. A key to achieving this vision is a collection of accurate time-resolved and spatially defined metabolite abundance data and associated metadata. One formidable challenge associated with metabolite profiling is the complexity and analytical limits associated with comprehensively determining the metabolome of an organism. Further, for metabolomics data to be efficiently used by the research community, it must be curated in publically available metabolomics databases. Such databases require clear, consistent formats, easy access to data and metadata, data download, and accessible computational tools to integrate genome system-scale datasets. Although transcriptomics and proteomics integrate the linear predictive power of the genome, the metabolome represents the nonlinear, final biochemical products of the genome, which results from the intricate system(s) that regulate genome expression. For example, the relationship of metabolomics data to the metabolic network is confounded by redundant connections between metabolites and gene-products. However, connections among metabolites are predictable through the rules of chemistry. Therefore, enhancing the ability to integrate the metabolome with anchor-points in the transcriptome and proteome will enhance the predictive power of genomics data. We detail a public database repository for metabolomics, tools and approaches for statistical analysis of metabolomics data, and methods for integrating these dataset with transcriptomic data to create hypotheses concerning specialized metabolism that generates the diversity in natural product chemistry. We discuss the importance of close collaborations among biologists, chemists, computer scientists and statisticians throughout the development of such integrated metabolism-centric databases and software.

In addition to neurodevelopmental effects, alcohol consumption at high levels during pregnancy is associated with immunomodulation and premature birth. Premature birth, in turn, is associated with increased susceptibility to various infectious agents such as Respiratory Syncytial Virus (RSV). The initial line of pulmonary innate defense includes the mucociliary apparatus, which expels microorganisms trapped within the airway secretions. Surfactant proteins A and D (SP-A and SP-D, respectively) are additional components of pulmonary innate immunity and have an important role in pulmonary defense against inhaled pathogens. The purpose of this study was to determine if chronic alcohol consumption during the third trimester of pregnancy alters the function of the mucociliary apparatus and expression of SP-A and SP-D of fetal lung epithelia. Sixteen, date-mated ewes were assigned to two different groups; an ethanol exposed group in which ewes received ethanol through surgically implanted intra-abomasal cannula during the third trimester of pregnancy, and a control group in which ewes received the equivalent amount of water instead of ethanol. Within these two groups, ewes were further randomly assigned to a full-term group in which the lambs were naturally delivered, and a pre-term group in which the lambs were delivered prematurely via an abdominal incision and uterotomy. Ethanol was administered 5 times a week as a 40% solution at 1gr/kg of body weight. The mean maternal serum alcohol concentration (SAC) measured 6 hr post administration was 16.3 +/− 4.36 mg/dL. Tracheas from 6 full-term lambs were collected to assess ciliary beat frequency (CBF). The lung tissue from all (24) lambs was collected for immunohistochemistry (IHC) analysis of SP-A and SP-D protein production and fluorogenic realtime quantitative polymerase chain reaction analysis (qPCR) of SP-A and SP-D mRNA levels. Exposure to ethanol during pregnancy significantly blocked stimulated increase in CBF though ethanol-mediated desensitization of cAMP-dependant protein kinase (PKA). In addition, pre-term born/ethanol-exposed lambs showed significantly decreased SP-A m-RNA expression when compared to the pre-term born/control group (p=0.004); no significant changes were seen with SP-D. The full-term/ethanol exposed lambs had no significant alterations in mRNA levels, but had significantly less detectable SP-A protein when compared to the full-term/control lambs (p=0.02).Corresponding author: Tatjana Lazic, DVM, MS, 2740 Veterinary Medicine, Department of Veterinary Pathology, Iowa State University, Ames, Iowa 50011-1250, 515-294-6688, FAX 515-294-5423, tlazic@iastate.edu. Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process error...

Network-based analysis is indispensable in analyzing high throughput biological data. Based on the assumption that the variation of gene interactions under given biological conditions could be better interpreted in the context of a large-scale and wide variety of developmental, tissue, and disease, we leverage the large quantity of publicly-available transcriptomic data > 40,000 HG U133A Affymetrix microarray chips stored in ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) using MetaOmGraph (http://metnet.vrac.iastate.edu/MetNet_MetaOmGraph.htm). From this data, 18,637 chips encompassing over 500 experiments containing high quality data (18637Hu-dataset) were used to create a globally stable gene co-expression network (18637Hu-co-expression-network). Regulons, groups of highly and consistently co-expressed genes, were obtained by partitioning the 18637Hu-co-expression-network using an MCL clustering algorithm. The regulon were demonstrated to be statistically significant using a gene ontology (GO) term overrepresentation test combined with evaluation of the effects of gene permutations. The regulons include approximately 12% of human genes, interconnected by 31,471 correlations. All network data and metadata is publically available (http://metnet.vrac.iastate.edu/MetNet_MetaOmGraph.htm). Text mining of these metadata, GO term overrepresentation analysis, and statistical analysis of transcriptomic experiments across multiple environmental, tissue, and disease conditions, has revealed novel fingerprints distinguishing central nervous system (CNS)-related conditions. This study demonstrates the value of mega-scale network-based analysis for biologists to further refine transcriptomic data derived from a particular condition, to study the global relationships between genes and diseases, and to develop hypotheses that can inform future research.

BackgroundThe synthesis of information across microarray studies has been performed by combining statistical results of individual studies (as in a mosaic), or by combining data from multiple studies into a large pool to be analyzed as a single data set (as in a melting pot of data). Specific issues relating to data heterogeneity across microarray studies, such as differences within and between labs or differences among experimental conditions, could lead to equivocal results in a melting pot approach.ResultsWe applied statistical theory to determine the specific effect of different means and heteroskedasticity across 19 groups of microarray data on the sign and magnitude of gene-to-gene Pearson correlation coefficients obtained from the pool of 19 groups. We quantified the biases of the pooled coefficients and compared them to the biases of correlations estimated by an effect-size model. Mean differences across the 19 groups were the main factor determining the magnitude and sign of the pooled coefficients, which showed largest values of bias as they approached ±1. Only heteroskedasticity across the pool of 19 groups resulted in less efficient estimations of correlations than did a classical meta-analysis approach of combining correlation coefficients. These results were corroborated by simulation studies involving either mean differences or heteroskedasticity across a pool of N > 2 groups.ConclusionsThe combination of statistical results is best suited for synthesizing the correlation between expression profiles of a gene pair across several microarray studies.