Challenges for 'omics' Technologies
Recent advances in applying mass spectroscopic methods to biological molecules and metabolites represents a paradigm shift for the life sciences. Mass spectroscopic methods are being used in conjunction with proteomic, genomic, and informatic techniques to identify and measure multiple analytes in a highly parallel fashion. These methods will ultimately provide insights into how cells regulate gene expression in response to their environment. The knowledge derived from metabolomics will ultimately improve our quality of life by providing better therapeutics, healthier foods, and a cleaner and safer environment.
The three main challenges facing metabolomic research today are: 1) instrumentation, 2) data acquisition and storage and 3) data analysis and interpretation. In terms of data analysis and interpretation, the specific challenge will be to transform large databases into information and information into knowledge. This will necessitate the development of novel biocomputation tools and theories that can reveal the non-linear principal dimensions imbedded in complex, high-dimensional data sets. Click on the figure to enlarge it
The Bioinformatics Shared Resource Core (BSRC),
a unit of the NCMHD Center of Excellence for Nutritional Genomics has recently analyzed a data set of 133 SELDI-TOF spectra from blood samples taken from individuals with and without prostate cancer. Figure 1a shows a portion of the "raw" data from a SELDI-TOF spectrum. Figure 1b, shows the same spectrum after filtering with a signal processing algorithm and application of a peak-finding algorithm.
Figure 2 shows a 3-dimensional model of the 133 spectra created using the novel non-linear algorithm, ISOMAP. This figure reveals that the first principal underlying dimension for this data set is the difference between normal (blue) and diseased (red) samples. This example demonstrates the power of the ISOMAP algorithm for reducing large, complex data sets to biologically relevant information. ISOMAP is one of about six novel non-linear approaches we are evaluating and the one with which we have had the most experience to date.
Dimensionality reduction of large complex data sets will be essential for the success of the metabolomic approach. The BSRC has the expertise and experience to address multiple dimensionality and it is eager to collaborate with scientists working in the exciting new field of metabolomics.
German, JB, Dillard, CJ, Hillyard, SL, Lange, MC, Smilowitz, JT, Ward, RE, Zivkovic, AM. 2006. Metabolomics: Bringing Nutrigenomics to Practice in Individualized Health Assessment. In Nutritional Genomics: Discovering the Path to Personalized Nutrition. Kaput, J and Rodriguz, R (eds). Wiley and Sons, Inc. NY. 2006.
Tenenbaum, JB, de Silva, V, Langford, JC. 2000. A global geometric framework for nonlinear dimensionality reduction. Science 290, 2319-2323. PMID: 11125149