A knowledge-based clustering algorithm driven by Gene Ontology.

We have developed an algorithm for inferring the degree of similarity between genes by using the graph-based structure of Gene Ontology (GO). We applied this knowledge-based similarity metric to a clique-finding algorithm for detecting sets of related genes with biological classifications. We also combined it with an expression-based distance metric to produce a co-cluster analysis, which accentuates genes with both similar expression profiles and similar biological characteristics and identifies gene clusters that are more stable and biologically meaningful. These algorithms are demonstrated in the analysis of MPRO cell differentiation time series experiments.

NetAffx Gene Ontology Mining Tool: a visual approach for microarray data analysis.

SUMMARY: The NetAffx Gene Ontology (GO) Mining Tool is a web-based, interactive tool that permits traversal of the GO graph in the context of microarray data. It accepts a list of Affymetrix probe sets and renders a GO graph as a heat map colored according to significance measurements. The rendered graph is interactive, with nodes linked to public web sites and to lists of the relevant probe sets. The GO Mining Tool provides visualization combining biological annotation with expression data, encompassing thousands of genes in one interactive view. AVAILABILITY: GO Mining Tool is freely available at http://www.affymetrix.com/analysis/query/go_analysis.affx

Gene structure-based splice variant deconvolution using a microarray platform.

MOTIVATION: Alternative splicing allows a single gene to generate multiple mRNAs, which can be translated into functionally and structurally diverse proteins. One gene can have multiple variants coexisting at different concentrations. Estimating the relative abundance of each variant is important for the study of underlying biological function. Microarrays are standard tools that measure gene expression. But most design and analysis has not accounted for splice variants. Thus splice variant-specific chip designs and analysis algorithms are needed for accurate gene expression profiling. RESULTS: Inspired by Li and Wong (2001), we developed a gene structure-based algorithm to determine the relative abundance of known splice variants. Probe intensities are modeled across multiple experiments using gene structures as constraints. Model parameters are obtained through a maximum likelihood estimation (MLE) process/framework. The algorithm produces the relative concentration of each variant, as well as an affinity term associated with each probe. Validation of the algorithm is performed by a set of controlled spike experiments as well as endogenous tissue samples using a human splice variant array.

GPCR-GRAPA-LIB--a refined library of hidden Markov Models for annotating GPCRs.

GPCR-GRAPA-LIB is a library of HMMs describing G protein coupled receptor families. These families are initially defined by class of receptor ligand, with divergent families divided into subfamilies using phylogenic analysis and knowledge of GPCR function. Protein sequences are applied to the models with the GRAPA curve-based selection criteria. RefSeq sequences for Homo sapiens, Drosophila melanogaster, and Caenorhabditis elegans have been annotated using this approach.

NetAffx: Affymetrix probesets and annotations.

NetAffx (http://www.affymetrix.com) details and annotates probesets on Affymetrix GeneChip microarrays. These annotations include (i) static information specific to the probeset composition; (ii) sequence annotations extracted from public databases; and (iii) protein sequence-level annotations derived from public domain programs, as well as libraries of hidden Markov models (HMMs) developed at Affymetrix. For each probeset, NetAffx lists the probe sequences, and the consensus sequence interrogated by the probes; for the larger chip sets, interactive maps display this sequence data in genomic context. Sequence annotations include Gene Ontology (GO) terms and depiction of GO graph relationships; predicted protein domains and motifs; orthologous sequences; links to relevant pathways; and links to public databases including UniGene, LocusLink, SWISS-PROT and OMIM.

Exploring alternative transcript structure in the human genome using blocks and InterPro.

Understanding how alternative splicing affects gene function is an important challenge facing modern-day molecular biology. Using homology-based, protein sequence analysis methods, it should be possible to investigate how transcript diversity impacts protein function. To test this, high-quality exon-intron structures were deduced for over 8000 human genes, including over 1300 (17 percent) that produce multiple transcript variants. A data mining technique (DiffMotif) was developed to identify genes in which transcript variation coincides with changes in conserved motifs between variants. Applying this method, we found that 30 percent of the multi-variant genes in our test set exhibited a differential profile of conserved InterPro and/or BLOCKS motifs across different mRNA variants. To investigate these, a visualization tool (ProtAnnot) that displays amino acid motifs in the context of genomic sequence was developed. Using this tool, genes revealed by the DiffMotif method were analyzed, and when possible, hypotheses regarding the potential role of alternative transcript structure in modulating gene function were developed. Examples of these, including: MEOX1, a homeobox-containing protein; AIRE, involved in auto-immune disease; PLAT, tissue type plasminogen activator; and CD79b, a component of the B-cell receptor complex, are presented. These results demonstrate that amino acid motif databases like BLOCKS and InterPro are useful tools for investigating how alternative transcript structure affects gene function.

Structure-based comparison of four eukaryotic genomes.

The field of comparative genomics allows us to elucidate the molecular mechanisms necessary for the machinery of an organism by contrasting its genome against those of other organisms. In this paper, we contrast the genome of homo sapiens against C. Elegans, Drosophila melanogaster, and S. cerevisiae to gain insights on what structural domains are present in each organism. Previous work has assessed this using sequence-based homology recognition systems such as Pfam [1] and Interpro [2]. Here, we pursue a structure-based assessment, analyzing genomes according to domains in the SCOP structural domain dictionary. Compared to other eukaryotic genomes, we observe additional domains in the human genome relating to signal transduction, immune response, transport, and certain enzymes. Compared to the metazoan genomes, the yeast genome shows an absence of domains relating to immune response, cell-cell interactions, and cell signaling.

Protein-based analysis of alternative splicing in the human genome.

Understanding the functional significance of alternative splicing and other mechanisms that generate RNA transcript diversity is an important challenge facing modern-day molecular biology. Using homology-based, protein sequence analysis methods, it should be possible to investigate how transcript diversity impacts protein structure and function. To test this, a data mining technique ("DiffHit") was developed to identify and catalog genes producing protein isoforms which exhibit distinct profiles of conserved protein motifs. We found that out of a test set of over 1,300 alternatively spliced genes with solved genomic structure, over 30% exhibited a differential profile of conserved InterPro and/or Blocks protein motifs across distinct isoforms. These results suggest that motif databases such as Blocks and InterPro are potentially useful tools for investigating how alternative transcript structure affects gene function.