Gene expression measurement using cDNA and oligo arrays is exploding in popularity, yet many technical problems remain. In parallel, high throughput methods for measuring protein concentrations are being developed. One of the more challenging problems results from the large volume of data generated in these experiments. Image capture, processing, interpretation and quantification remain as important fundamental issues. Quality control and experimental design must be carefully addressed. Many statistical, image processing and bioinformatics issues remain problematic. Accordingly, these projects seek to address such problems. One of the major remaining technical problems with interpretation of Affymetrix microarray data is in the computation of the proper gene expression index from the raw chip intensity data. We have proposed and recently enhanced a statistical data transformation (""""""""symmetric adaptive transform"""""""") which allows data to be presented on a statistically optimal, uniform variance scale. This transform has been incorporated into the MSCL Analyst's Toolbox. In a major NIH-wide project, we maintain a database for storage, retrieval and analysis of Affymetrix oligo-based microarrays, NIHLIMS. As part of this collaboration, we are creating a data analysis pipeline and bioinformatics toolset, including both commercial and freely available software. The database currently stores information from over 2000 microarrays. Our web-based analysis (A-SCAN) and downloadable tool set (MSCL Analyst's Toolbox) are now mature, widely tested and applied in numerous studies. Working with laboratories in NCI, CC, NHLBI, NINDS, NIAID, NHGRI, NICHD, NIA, NIDDK, NIDA we have developed, customized and applied software for the analysis of microarray based studies. We also maintain a quarterly-updated set of annotation files for use with Affymetrix data, in a format for convenient download and use by our collaborators. We are also performing as the """"""""analysis core"""""""" for a high-volume microarray laboratory in CCMD/CC. All microarray studies by this group now pass through our analysis pipeline. In one such study, a novel, mitochondrial gene specific chip developed by a group in NINDS is being compared to standard Affymetrix chips to determine whether the performance characteristics of the custom chip are satisfactory. The chip assays a unique set of mitochondrial genes, not interrogated by other chips, which focus rather on nuclear genes. The novel chip design required a new statisical procedure for normalization and analysis. In a new series of studies with investigators in NIDCR, we are currently analyzing gene expression in human monocytes before and after differentiation into dendritic cells, under stimulation by lipopolysaccharides derived from bacteria of interest to dental research. The goal is to reveal any basic differences in host response to different organisms, which may be useful diagnostically or therapeutically. In a large proteomics initiatives, we collaborated with investigators in CCMD to apply mass-spectroscopy for identification of novel entities as possible markers of infection in BAL (bronchial alveolar lavage) fluid, blood serum and plasma. A time course study of normal volunteers exposed to endotoxin demonstrated a significant number of varying spectroscopic peaks, which were subsequently identified molecularly. A study of samples from ARDS (acute respiratory distress syndrome) patients confirmed most of these results. We have developed improved software for applying statistical tests to entire mass-spectra, to objectively determine which peaks yield discriminatory information. We have recently applied modern statistical algorithms, including Random Forests, to ascertain whether combinations of peaks could improve diagnostic discrimination, over using single peaks. In a study of pheochromocytoma, with investigators in NHGRI, NICHD, NINDS and other centers, we have applied novel statistical techniques to refine the analysis of 2 color, oligonucleotide printed microarrays. Having recognized the large """"""""batch"""""""" effect present in an earlier study by this same group, we proposed effective analytical solutions which minimized this contaminating effect and allowed for improved statistical power to detect potential markers of malignant tumors.
Showing the most recent 10 out of 46 publications
