This page contains supplementary material for our High-Throughput GoMiner paper.
Fresh blood was obtained from the patient and six age- and sex-matched healthy donors. Peripheral blood mononuclear cells (PBMC) were purified and then stimulated with anti-CD3-coated and anti-CD28-coated beads for 24 hours. Total RNA was extracted from the PBMC samples with Qiagen�s RNeasy mini Kit.
mRNA was then amplified into cRNA using T7 RNA polymerase , in the process incorporating amino-ally5-(3-aminoallyl)-UTP. After synthesis, the cRNA was labeled with Cy3 or Cy5 fluorescent dye by coupling to the derivatized UTP. 70-mer Oligonucleotides representing 14,000 genes were synthesized by Operon , and the microarrays were printed in the NIH NIAID Microarray Center.
In each experiment, the cRNA sample from a patient or normal control was labeled with Cy5 and then hybridized to the oligonucleotide microarray, along with a Cy3-labelled mRNA sample prepared from a pool of PBMC stimulated with PMA and ionomycin. The pooled mRNA provided a common reference for comparing the relative expression of each gene across all samples.
Hybridized microarrays were scanned with a GenePix 4000 microarray scanner (Axon Instruments) to obtain fluorescence images. The images were analyzed with GenePix Pro software, and the ratios of intensities in the red and green channels were calculated. Low quality fluorescent spots were flagged and excluded from subsequent analyses.
The raw data were uploaded into a custom microarray database, mAdb developed by J. Powell, et al. . Fluorescence ratios were normalized for each microarray by applying a scaling factor so that the median fluorescence ratio of well-measured spots on each microarray was 1.0. All non-flagged microarray elements with fluorescence intensity in each channel greater than two times the local background were used. Genes with adequate data from more than 60% of the microarrays were included in the further analysis. 4152 genes met those criteria. For each biological sample, an average value of gene expression was taken over all replicates. A novel a posteriori high-throughput method (Zeeberg and Qin, in preparation) was used to select the �optimal� a posteriori candidate set of fifty-seven genes, each of which showed more than a three-fold difference in expression level relative to at least 5 of 6 normal controls.
A file containing the HUGO names of the 3968 unique genes detected on the microarrays and a second file containing the HUGO names of the 57 genes showing three-fold or more differential expression (Supplementary Materials section �Expression Data�) were used as input to High-Throughput GoMiner. After processing, the summary report output file (Figure 2 - see paper; Supplementary Materials section �Summary Report�) indicated that 30 Gene Ontology (GO) categories showed an FDR of less than 0.10 and that 24 of the 57 changed genes were present in those categories.
Because of the parent-child structure of the Gene Ontology, some of the 30 categories are likely to be similar, or even identical, to one another in terms of the changed genes that are members. To evaluate that possibility and to determine the reduced number of truly different categorical clusterings, we used the CIMminer program  to create a clustered image map (CIM) visualization of categories versus changed genes (Figure 3 - see paper).
The GO categories were clustered on the basis of co-occurrence of the differentially expressed genes (centered Pearson correlation metric; average linkage clustering). Yellow indicates that the gene is not present in the category. Red and green indicate over- and under-expression, respectively, compared with the normal controls. Clusters of identical or very similar categories can be derived from the CIM and are shown in Table 1 (see paper). The original 30 categories appear to fall into seven clusters. This analysis represents a novel use of the CIM paradigm.
GoMiner™ is a development of the Genomics and Pharmacology Facility, Developmental Therapeutics Branch (DTB), Center for Cancer Research (CCR), National Cancer Institute (NCI).
Please email us with any problems, questions or feedback on the tool.
Notice and Disclaimer