Welcome to the homepage of the Bioinformatics and Computational Biology (BICB) microarray group at UCSD. We have developed a collection of Java-based tools for the interpretation of functional genomics data. Each of these tools addresses a different aspect of microarray statistics. VAMPIRE and BMA can be used to identify statistically significant differences in gene expression between treatment conditions. GOby can be used to identify enrichment of functions, processes, components and pathways in a list of "selected" genes.
VAMPIRE was originally developed to interpret one-channel microarray data, such as Affymetrix oligonucleotide arrays. Given a summary measure of gene expression, such as the Affymetrix MAS 5.0 scores for each microarray feature (or probe set), it determines the optimal variance model parameters for a two-component variance model. The expression-independent variance represents a constant "background" noise that affects all array features to the same extent, while the expression-dependent variance represents a proportional noise that increases with gene expression. Low-intensity features thus have larger proportional of noise, because of the influence of expression-independent variance. With this optimized model, VAMPIRE then computes a Bayesian statistical test to determine whether observed changes in intensity are statistically significant. For more background reading, the VAMPIRE paper can be found in Bioinformatics.
BMA was developed for the interpretation of two-channel microarrays, such as the Aglient oligonucleotide arrays. In these experiments, two RNA samples are labeled with different dyes and hybridized to the array. In order to overcome dye bias, RNA samples are also measured with their dyes reversed. With this dye-pair, BMA can determine which genes are differentially-expressed between the two RNA samples. BMA works by modeling not only expression-independent and expression-dependent variability, but by also modeling the degree of correlation between the paired measurements on a single array. BMA then uses a Bayesian statistical test based on the bivariate normal distribution to determine the significance of observed differences in expression. The manuscript for this approach is currently in review.
GOby is a tool for beginning biological interpretation of differentially-regulated genes. It identifies Gene Ontology (GO) terms or KEGG pathways that are significantly enriched in a list of "selected" genes. By comparing the gene annotations of "selected" genes with the gene annotations of "background" genes, we can determine whether a particular GO term is enriched. This problem is essentially the "urn" problem, which is solved by computing the appropriate value from a hypergeometric distribution. More details about this tool have been published with the web-based VAMPIRE interface in Nucleic Acids Research.
There are a number of additional tools in the works, and suggestions for new features are always welcome. Feel free to contact the relevant authors.