Toxicogenomics through the Eyes of Informatics: conference overview and recommendations.

Virginia Bioinformatics Institute, in conjunction with National Institutes of Environmental Health Sciences, hosted a conference, "Toxicogenomics through the Eyes of Informatics," in Bethesda, Maryland, USA, on 12-13 May 2003. Researchers around the world met to discuss how the application of bioinformatics tools, methodologies, and technologies will enhance our understanding of how cells and organisms respond to toxins. Conference topics included statistical methods, quantitative molecular data sets, computational algorithms for data analysis, computational modeling and simulation, challenges and opportunities in computational biology, and information technology infrastructure for data and tool management. This meeting report is a summary of conference presentations, survey results, current toxicogenomics concerns, and future directions of the toxicogenomics community. In conclusion this report discusses toxicogenomics as related to environmental agents, cell-chemical reactions, and gene-environment interactions. Key words: bioinformatics, cell-chemical reactions, environmental agents, toxicogenomics, toxins, Virginia Bioinformatics Institute.

**********

The availability of genome-scale DNA sequence information has radically altered life science research. Moreover, the human genome project has revolutionized the practice of biology and the potential of medicine. The field of toxicology will be greatly augmented by the application of the knowledge of genetics and its sister sciences. Consequently, researchers have entered a new area of specialty--toxicogenomics--the marriage of toxicology and genomics. Data generated by such research will impact many areas including health and environmental sciences. In addition, the science and health care communities will also be better able to identify a wide range of human health risks. However, to realize the potential of toxicogenomics, a global view of gene function and temporal and spatial regulations for all genes at different physiological or pathological states or developmental stages is paramount. The expanding field of bioinformatics offers the convergence of biological, mathematical, and computational platforms to better understand the effects of toxins on biological systems.

Scientists using data developed by the modern advent of biotechnology need novel and efficient bioinformatics tools to monitor gene-environment interactions on a large scale. These bioinformatics platforms (hardware, software, portals, and models, for example) will allow the conversion of data into useful knowledge to prevent disease resulting from environmental exposures.

Conference Overview

The conference "Toxicogenomics through the Eyes of Informatics" was held 12-13 May 2003 in Bethesda, Maryland, USA. The purpose of the conference was to bring together researchers around the world to discuss the application of bioinformatics tools, methodologies, and technologies in order to increase our understanding of how cells and organisms respond to toxins. By understanding the mechanisms from a genomic and a systems view, investigators can more rationally assess and evaluate the risk that toxins pose to humans. The format of this meeting aimed to facilitate input from the range of disciplines that can now provide insight into the field. Table 1 illustrates the conference program with speakers and events. By integrating experimental studies at the structural, molecular, and cellular levels with mathematical and biochemical modeling and informatics, the conference surveyed the integration and merger of advancing technologies. This conference provided a forum to explore the potential applications and implications of the new technologies, centering on toxicogenomics research platforms.

Recommendations and Survey Report

Researchers completed a survey about the future and importance of toxicogenomics (Table 2). The presentations and survey responses are summarized in the following observations:

* The complex characteristics of biological data generated by bioinformatics approaches (genomics, transcriptomics, metabolomics, proteomics) offer an opportunity to explore data analysis and standardization. Furthermore, data analysis via mathematical and biochemical modeling is paramount in predicting the effects of toxins on biological systems.

* Novel visualization tools will be required for researchers to realize the usefulness of bioinformatics data in toxicogenomics research, For investigators to catalyze forward movement in these new directions, however, baseline (or normal) data for systems, including normal ranges of variation, must be generated and centrally stored in publicly available databases.

* To study the effects of toxins on biological organisms holistically, we must develop more sophisticated and robust computational methods and bioinformatics infrastructures for biological knowledge discovery, including novel [information technology (IT)] and database applications. In addition, predictive modeling will allow for greater understanding and, in turn, knowledge extraction from metabolic and proteomic profiles currently being generated.

* To develop and establish coordinated database and informatics platforms (e.g., portals) for data exchange and analysis, we need to define primary end-user needs and major issues facing relevant subdisciplines. The computational and networking needs for toxicogenomics research should be defined. Moreover, the system must integrate with similar efforts in other health-related disciplines to merge all areas of the life sciences in a scalable, flexible system. This effort will depend on exploring data standards and system interoperability concurrently. The development of toxicogenomics data generation center(s) of excellence linked via bioinformatics center(s) of excellence would allow the compilation, analysis, and sharing of relevant data and tools.

* Toxicogenomics research programs should be developed to facilitate interoperability among the varying activities under way. Each individual research project will be more successful, ultimately, if it is part of a larger coordinated and synergistic approach.

* Mathematical modeling will continue to reveal qualitative patterns and define crucial parameters of biological systems. These efforts will facilitate the design of experiments and increase our understanding of complex biological processes. The process is spiral in that models suggest new hypotheses to be tested in new experiments, but the latter will provide data to improve the models. At each turn knowledge increases and uncertainty decreases.

* Several reports concur that the demand for qualified bioinformaticians far outstrips the current number of qualified graduates. The development of multidisciplinary educational and training programs in bioinformatics, and training for toxicogenomics research in particular, must be in step with scientific discovery.

* Programs to support additional toxicogenomics conferences and workshops would allow communities to foster communication and coordinate efforts.

Conclusion

Bioinformatics tools for disease research will catalyze our quest to understand how individuals differ in their susceptibility to environmental agents and how these susceptibilities change over time. As the inherent goal of systems biology continues to intertwine experimentation, mathematical modeling, and computation, enhancing the investment in each of these efforts will hasten our understanding of cell-chemical interactions. At this level, functional, structural, and comparative analysis help provide insights into the environmental influence predisposing plants, animals, and humans to disease. Recent contributions of bioinformatics research have yielded molecular data on expressed genes (RNA), proteins, and metabolites. Unlike genomic sequencing (DNA), proteomic, metabolomic, and transcriptomic data provide "state" information that helps form a dynamic description of how the organism interacts with its environment. Age, developmental stage, general health, and nutritional status have significant influence on disease outbreaks, whether caused by environmental pollution, toxic agents, or pathogens. An integrated approach to exposure assessment, epidemiology, and data collection, for example, would allow quicker identification of susceptibility genes and characterization of their functions. If bioinformatics data relating to human disease, regardless of cause, could be integrated and manipulated via bioinformatics platforms available to all research stakeholders, the scientific community could achieve an integrated view of gene-environment interaction at the level of whole organisms.

This resulting comprehensive catalog of human gene variation and response would increase our understanding of chronic disease. Subsequently, our capability to control environmental toxin levels and pathogens that cause disease to plants, animals, and humans would increase. However, many of the clinical, scientific, and pathogen data generated today are not put into a meaningful and integrated context that would enable problem solving via bioinformatics applications. Limited bioinformatics platforms allow biological annotations to be easily integrated with related host-baseline or broader environment or ecosystem data when available. A centrally funded program in bioinformatics research dedicated to the synthesis and informatics to support toxicogenomics research would provide for the realization of unprecedented opportunities to elucidate biological responses to environmental toxicants and stressors at the genomic level.

Table 1. Toxicogenomics through the Eyes of Informatics:
conference program and presentations.

Session/speaker         Affiliation             Title of presentation

IT infrastructure

Barbara Eckman          IBM Life Sciences       Data integration and
                                                management for
                                                molecular and cell
                                                biology

Rolf Apweiler           European                Integration and
                        Bioinformatics          standardization:
                        Institute               driving forces in the
                                                development of database
                                                resources

Shankar Subramaniam     University of           Bioinformatics
                        California San Diego    challenges in the
                                                post-genomic era

Computational
modeling and
simulation

Zoltan Oltvai           Northwestern            Organization of
                        University              cellular networks

Pedro Mendes            Virginia                Inferring biochemical
                        Bioinformatics          networks from
                        Institute               functional genomic data

Hiroaki Kitano                                  Luncheon address

Reinhard Laubenbacher   Virginia                Reverse-engineering in
                        Bioinformatics          computational
                        Institute               immunology

Herbert Sauro           Keck Graduate           Data exchange and
                        Institute               programmatic resource
                                                sharing: the Systems
                                                Biology Workbench
                                                (SBW), bioSPICE, and
                                                the Systems Biology
                                                Markup language

John Tyson              Virginia Polytechnic    From molecular
                        Institute and State     regulatory networks to
                        University              cell physiology

Charles deLisi          Boston University       Inferring the function
                                                of genes and networks

Statistical methods
and quantitative
molecular data sets

Bruce Weir              North Carolina State    Association patterns in
                        University              the human genome

Lloyd Sumner            Noble Foundation        Proteomics and
                                                metabolomics: temporal
                                                snapshots of the
                                                biochemistry of
                                                Medicago truncatula

Carolyn Mattingly       Mount Desert Island     The Comparative
                        Biological Laboratory   Toxicogenomics Database
                                                (CTD): new perspectives
                                                through data
                                                integration and
                                                curation

John Quackenbush        The Institute for       Interpreting expression
                        Genomic Research        through biology

Ute Roessner-Tunali     Max-Planck-Institute    Contrasting and
                        for Molecular Plant     combining information
                        Physiology              derived from the
                                                systems biology
                                                approaches of
                                                metabolite and
                                                transcript profiling

Computational
algorithms for data
analysis and
visualization

Heike Hofmann           Iowa State University   Graphical opportunities
                                                in exploring microarray
                                                data

Lior Pachter            University of           Comparative genomics
                        California Berkeley     by phylogenetic
                                                shadowing

Tarif Awad              Affymetrix, Inc.        Gene ontology graphs as
                                                mining and display
                                                tools for gene
                                                expression data

Keith Allen             Paradigm Genetics       Integrating metabolomic
                                                and gene expression
                                                data

Computational
biology: looking
forward

Stefan Unger            Sun Microsystems        Computational biology:
                                                challenges and
                                                opportunities

Congressman Virgil H.   Fifth District of       Passing the hat
Goode Jr.               Virginia

Table 2. Conference participant questionnaire.

In your view, what are critical areas of research--that if enhanced
via additional funding--would enable the integration and use of
mathematics, modeling, simulation, and IT infrastructure to enhance
our understanding of cell-chemical interactions?

How do you think experimental approaches will be altered or enhanced
via advances in mathematics, modeling, simulation, and IT
infrastructure? For example, can experimental designs be altered to
maximize data set utility?

How might mathematics, modeling, simulation, and IT infrastructure
alter laboratory, macrostructures, and "omics" data generation and
vice versa?

What do experimentalists require most of mathematics, modeling,
simulation, and IT infrastructure to remove key bottlenecks?

Kenneth Olden, (1) Neysa Call, (2) Bruno Sobral, (2) and Robin Oakes (2)

(1) National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina, USA; (2) Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA

Address correspondence to B, Sobral, Virginia Bioinformatics Institute, Virginia Tech, Bioinformatics Facility 1 (0477), Blacksburg, VA 24061 USA. Telephone: [phone], Fax: [phone]. E-mail: [e-mail]

We thank J. Selkirk from the National Institutes of Environmental Health Sciences for his scientific guidance. We also thank C. Phillips from Virginia Bioinformatics Institute for her assistance with conference arrangements and logistics.

The authors declare they have no competing financial interests.

Received 14 October 2003; accepted 31 March 2004.