We have adapted Gibbs sampling strategy, which has become a method of choice for the discovery of motifs in DNA and protein sequences, to the biclustering of discretized microarray data. In contrast with standard clustering, biclustering reveals similar expressional behavior of the genes over a subset of conditions in an microarray data set.

Hidden Multivariate Markov Models (HM³s) are introduced for modeling multi-dimensional genomic DNA data. A bivariate version of HM³s is developed for studying the joint behavior of the C+G richness pattern and the bendability pattern of DNA. Applications of the bivariate HM³s for recognition/prediction of eukaryotic promoter regions are illustrated.

We introduce a new method for detecting statistically meaningful functional associations between domains from molecular complexes. Two random control sets were used to compute P-values for domain co-occurrences in complexes. Results from four different datasets show that many of the correlations are between domains of similar or associated function.

ReBIL aims to improve the efficiency of information extraction systems applied to biological literature. The project is based on the correlation between structural and functional classifications of gene products. We evaluate extracted information by checking if they preserve the correlation. More information about Rebil is available at http://xldb.fc.ul.pt/rebil/.

RNA polymerase II - although important as it transcribes all the protein coding genes in the cell, little is known about its termination process. This study focuses on identifying motifs that link to transcription termination and polymerase relase process.

GOstat provides a useful tool in order to find biological processes or annotations characteristic for a group of genes. This is greatly helpful in analyzing lists of genes resulting from high throughput screening experiments, such as microarrays, for their biological meaning. GOstat is accessible via the Internet at http://gostat.wehi.edu.au.

Multi-Dynamic Bayesian Networks (MDBNs) are introduced for analyzing multi-dimensional genomic DNA data. A two-dimensional version of MDBNs is developed for learning and predicting the joint behaviour of the C+G richness pattern and the bendability pattern of DNA. Applications of these MDBNs for recognition/prediction of eukaryotic promoter regions are illustrated.

We present GATO (gene annotation tool) a tool to analyse gene expression data based on Ensembl database and Gene Ontology. We describe how our tool can be utilized to rapidly mine gene expression data and drive biological interpretation using Affymetrix arrays.

We present an image analysis system to detect and count eukaryotic cells in darkfield microscopy images. Analyzing undiluted yeast suspension the tool differentiates between single, budding and cell clusters as well as dead and vital cells. The cells within a cluster are detected by active contours as well as the cell features which results in precise information for fermentation.

As the amount of genes with known function available is growing there is a need for classification methods which allow the use of prior knowledge. Partially supervised clustering of time courses stemming from gene expression experiments addresses this problem. Here a model-based clustering approach using Hidden Markov Models is proposed.

Extracting molecular interactions from the rapidly growing biomedical literature is important to seek systematical explorations of relationships between genes and proteins. However, many of the existing computer-aided methods are not sufficiently capable of processing a huge amount of literature. Extraction techniques include molecule name detection, interaction event detection, and graphical interface construction. We explore these techniques and show system examples.

With the rapid growth of machine-readable literature, such like MEDLINE database, a search for articles is an important assignment. Therefore, we propose an efficient algorithm to select information from results of a PubMed search. When taking 487 UV-regulated genes, it extracted sentences containing the query with 97% precision and 97% recall.

TextLens Partial Parser is a parser to catch a pair of main subject and predicate of a sentence. It aims at an improvement of information extraction by appropriately capturing each chunk of words and a structure of a sentence. It uses a rule-based algorithm which makes an abstract expression of a sentence.

MutationMiner is a program which automates extraction of point mutation data from biomedical literature. It uses regular expressions and a graph theoretic approach to find point mutations in the text and then confirms the mutations with SwissProt data. MutationMiner can search over one thousand journal articles in 24 hours.

We have used phylogenetic algorithms and hidden Markov models to identify functionally distinct subsets within the family of chromodomains. Our results demonstrate the validity of using bioinformatic analysis of large datasets to predict subtle but meaningful differences in protein domain function and structure-function relationships.

FatiGO (http://fatigo.bioinfo.cnio.es) is a simple but powerful procedure to extract Gene Ontology terms that result(upon the application of a statistical test) significantly over or under-represented in sets of genes within the context of a genome-scale experiments (DNA microarray, proteomics, etc.).

GENIA corpus 3.0p and 3.01, consisting of 2,000 MEDLINE abstracts, have been released with linguistically rich annotations including sentence boundaries, term boundaries, term classifications, semi-structured coordinated clauses, recovered ellipsis in terms, part-of-speech, etc. This poster is intended to provide the current status of the GENIA corpus.

The rapid increase in protein structures produced by structural-genomics projects will make it increasingly difficult to analyse and understand the distribution of proteins in fold space. We have applied a machine-learning strategy to automatically determine the structural principles describing 45 folds.

We present a prototype implementation of a system for extracting protein/gene interactions from biological literature which is motivated by the theory of Construction Grammar. CG provides a powerful framework for combining domain-specific terminology management with patterns incorporating generic linguistic structural constraints.

Prediction of gene function is one of the most challenging tasks in genomic science when there is no clear sequence similarity to annotated genes. Here we present a new method denominated Anticorrelation of Gene Presence to predict gene function. Using this method we identified four new genes involved in thiamin biosynthesis

We developed a method based on a statistical analysis of Expressed Sequence Tags (ESTs) to evaluate the positional clustering of differentially expressed genes. Human chromosomes 20, 21 and 22 were analysed with this method and show clusters of specifically expressed genes

A data integration generalist tool has been developed allowing simultaneous query in several data sources. The query is restricted to specific types such as project or metabolite name. The hits are displayed with links to the originating data source and an option to use the hit as the next query.

DUBs are cysteine proteases controlling the ubiquitination status of target proteins. Using genome mining tools, we have conducted searches for new human DUB members leading to the identification of 11 new sequences. Here we report on their in silico annotation and discuss the primary functional characterization of selected members.

Interpro and GPCRDB are two GPCR classification systems. Using data mining technique, we compared the two systems family by family at each level of the classification hierarchy and established mapping relation between them. We introduced a novel visualization tool that allows us to directly and easily compare them.

We describe the application of association rule discovery technique to find relevant relations between different genes attributes and experimental conditions in microarrays expression dataset. This method can be used to extract interesting and very diverse biological information. The method is implemented in EngeneTM software package that it is freely available upon request at http://www.engene.cnb.uam.es

We present the strategy used to extract the "in silico expression profiles" of the human and mouse genes (EST mining approache) and the data describing differentially expressed genes, disease related genes, and cluster of genes potentially involved in a common cellular function. Orthology gene expression comparison will also be presented.

We compared the sequence-derived representative dataset PDB_SELECT with the structural database SCOP. Some folds remain overrepresented in PDB_SELECT. After filtering, we obtain a subset of unique protein fold representatives: approximately ¼ of the original PDB_SELECT 25% list. We also discuss using unique representatives of SCOP folds as a representative dataset.

The ability to process natural language based information computationally is becoming a necessity for the geneticist. We present here an architecture for Bioretrieve, a computational tool for the management of the extremely large body of knowledge that exists about genes. Designed in a modular fashion and employing an open-source approach, it has advantages over existing monolithic systems.

We are using a machine-learning algorithm implemented in GO-KDS to complement SwissProt literature citation fields for each database entry. SEMA organizes this new relevant information, then builds a conceptual navigable map that is presented to the user as a flat or hyperbolic tree. This map allows redefining queries over the same database or over other information sources.

We describe the application of Non-negative Matrix Factorization (NMF) technique to reduce dimensionality and to find local patterns hidden in gene expression data sets and in the scientific literature. Results show the potential of this new machine learning technique to find relevant biological information.

We develop GENAW (GEnetic Network Analysis Workbench for microarray raw data) system that produces automatically the network from raw expression data. GENAW accepts various data formats of commercial tools and provides various visualization tools. We experimented with Yeast cell cycle data from Stanford university, our experiment was sufficiently reasonable.

We devised a novel approach to integrate rules induced from multi-class and unbalanced data sets; and to demonstrate its usefulness to multi-class protein fold classification which contains 700 examples for 27 SCOP folds. We showed that this approach increases the sensitivity of the classifiers and yielding more useful classifiers.

C.elegans microarray data is analysed by a novel nonmetric multidimensional scaling method that is maximally nonmetric. The genes are embeddable in 3D. Their annotations are consistent with their positions in this space. A method to compute the 3D coordinates directly from the microarray data is also developed.

A nonmetric multidimensional scaling analysis of the gene activity response of cell cycle-synchronized human fibroblasts to serum [Lyer et al. Science 283, 83-87 (1999)] automatically gives a ring-like gene arrangement along which the expression level peak rotates. This unambiguously demonstrates the power of a nonlinear data mining method.

Staying abreast of research on transcription factors (TFs) is currently a difficult task for biologists. We are building a system that will extract TF interactions from Medline abstracts automatically. To date, we have annotated a corpus for TF interactions and evaluated a number of component technologies.

Our plant gene index system based on stackPACK EST clustering with tissue categorization contains valuable information about 150,000 consensus sequences obtained from 9 principal plant model organisms. The information can be browsable, searchable, and downloadable with user-friendly web-interface at http://plant.pdrc.re.kr/new_korea/genepool/Plant/index.html

We propose a new gene selection method based on forward selection method in regression analysis. This method reduces redundant information about cancer that could be in the subset of selected genes. The result shows high accuracy of 90.3% for colon cancer data.

Collaborations in web based bioinformatics environment require intelligent supports to assist human computer interaction. BioInfoCallboratory is an agent assisted web based environment for supporting bioinformatics research. It facilitates sophisticated interactions such as: matchmaking based on common interests; internet spanning data mining for bioinformatics data; and event alert for interested parties.

We have developed a rapid, automated search strategy for the detection of contaminating sequence from putative novel pathogenic archaea in human EST sequence data. The system has general application to the detection of microbial pathogens and will be available at http://psychro.bioinformatics.unsw.edu.au.

We present the Gene Ontology Clusterer (GOC), which structurally classifies the GO based on pseudo-distances between comparable nodes in posets, in conjunction with scoring algorithms, to rank-order the GO nodes with respect to a set of requested genes. We will also share lessons we've learned about working with the GO.

We define two negatively correlated ideal gene vectors which represent the patterns of classes well and extract two significant gene subsets (SGSs) based on the similarity to two ideal genes. We train the neural network classifiers with SGSs and combine them. The ensemble classifier produces the best recognition rate-97.1% in Leukemia, 87.1% in Colon, and 92.0% in Lymphoma.

Development of research analysis pipelines often involves working with poorly understood data to answer questions that are, initially, simplistic. This poster overviews some strategies and best practices that may be employed in such work, including; handling & appraisal of the data, choice of appropriate thresholds, extrapolation, and checking for reasonableness.

Approximately 8,200 human cDNA libraries in dbEST were hierarchically classified in a hierarchical fashion in four gene expression categories – tissue, pathology, developmental stage, and sex. Web-based application for profiling gene expression using the resulting database is available at http://genome.ewha.ac.kr/EODB/.

In this study, we propose a method to detect explicit or implicit protein-protein interactions from text data. It was applied to the detection of interactions between yeast proteins. The result indicates that the putative interactions detected by the method can contain true and experimentally unidentified interactions.

Intra-molecular disulphide bonds (IDSB) are an important determinant of a protein’s 3D conformation. This work describes the differences in IDSB arrangements between disulphide-rich and -poor proteins. A naturally occurring partition of 25.2 residues / IDSB was used, revealing differences in PDB headers, SCOP folds, IDSB bonding patterns and loop lengths.

BioNLPro is a system for providing the base for a robust bio-text mining system. It is an intimately integrated NLP system consisting of the adapted core NLP modules reflecting the peculiarities of bio-text including a POS tagger, a biological term recognizer, a grammatical relation tagger based on chunking and a biological event extractor.

CSIRO Bioinformatics have developed an analysis methodology based on generalized linear models coupled with a specialized Bayesian variable selection technique. This methodology is capable of producing parsimonious predictors of; Classification targets, Numeric targets using Gaussian, Poisson or Gamma regressions or Survival targets using Cox's proportional hazards regression.

We propose a new method for detecting the plagiarism by exploiting the genomic sequence alignment. The system extracts linear sequences of keywords from the source code flow, and computes the local alignments to detect local similarity of original sources. The experimental results show this approach is more powerful than fingerprinting-matching.

The CBBC conducts research in computational biology ranging from comparative genomic sequence analysis, microarray and proteomic data analysis, and novel algorithms and software. The CBBC is developing a comparative analysis framework incorporating audit trailing of analysis, open source activities and distributed resource management. We present an overview of this framework.

Detecting local functional sequence patterns is suitable in GPCR sequence analyses. We attempted to extract patterns that are specific in GPCR subtypes. For extracting patterns, we applied different rules to each domain of three GPCR domains. Consequently, we obtained specific patterns of GPCR clans with high frequency and high specificity.

Pathway is a way to present the mechanism behind a biological phenomenon. We have developed methods to integrate pathway-related information. By querying this integrated database, users will be able to look up tissue specific pathways and discover possible cross talk between pathways. The query results can output in graphic form.

We have developed and implemented a library for a general Hidden Markov Model (GHMM) to assist in designing a topology and visualizing parameters for a HMM. The tool has been used in solving problems such as identification of circular permutation with Profile HMMs. GHMM and HMMEd is freely available at http://sourceforge.net/projects/ghmm/.

We present an efficient algorithm for detecting putative regulatory elements in the upstream sequences of genes, using expression data obtained from microarrays. We are able to find the optimal pattern, most correlated with the expression levels of the genes, in time linear in the total length of the upstream sequences.

We present a method of discovering useful patterns from DNA microarray experiment with large-scale multifactor design with no replication using genetic algorithm. Permutation test for the distance measures between observation and the pattern discovered statistically significant multifactor gene expression patterns with simple biological interpretations.

We propose a new method of quantitative trait loci (QTL) mapping using genetic algorithm (GA) with simulated EM algorithm. It detects QTL without highly organized experimental cross and is applicable to human genetics. Simulation studies showed high performance of our method in the cases not supported by traditional gene mappings.

We introduce a novel hierarchical-partitioning clustering algorithm for gene expression data analysis, which combines both features of hierarchical-based and partitioning-based clustering. Our algorithm performs a successive binary subdivision of the data into smaller and smaller partitions hierarchically, until no further splitting of a (parent) partition into two smaller (children) partitions is possible.

We propose a new clustering method with the following two features. First, this method exploits a new similarity measure based on distribution of gene expressions. Second, this method leverages the P-quasi complete linkage algorithm for describing clusters. The synergy of the two features provides more informative clustering than traditional ones.

We present a method for the extraction of minimal complete sets of exactly repeated sequences from genomes, and then extracting subsets with the potential for producing primary sequence-dependent RNA regulatory signals and regulatory networks, initially focused on intronic sequences, wherein we can examine clustering of matched sequences by functional groups. Results are presented for S.cerevisiae.

This poster presents DESCRIBER, a system that uses graphical models to represent relational data in computational genomics portals cf. myGrid, integrating descriptive data models for microarray data mining and extending the information retrieval capabilities of indices such as ResearchIndex. The objective is to provide collaborative filtering (CF) over data, metadata, source code (cf. OpenBio), and experimental documentation.

We present BNJ, an experimental Java-based software toolkit for learning network models of gene regulation from microarray data. We survey current research issues in learning the structure of graphical models, outline the components of BNJ used in modeling regulatory dynamics of S. cerevisiae, and present preliminary results and current research directions.

This project aims to create an analytical pipeline that links existing pattern discovery tools to allow automation of transcriptional element pattern predictions to be preformed simultaneously across multiple species directly from microarray experimental results. The system may be used as a means to map transcriptional pathways to macrophage specific regulatory networks.

We present a novel implementation of protein superfamily clustering using biomedical literature via the recently introduced information bottleneck method which shows good performance in document clustering. We test our method over 1866 saccharomyces cerevisiae proteins in COGs (Clusters of Orthologous Groups of proteins) by NCBI (National Center for Biotechnology Information).

We describe an algorithm that identifies families of genes with similar nucleotide patterns and hopefully similar function. The approach is quite general in terms of its flexibility with respect to the number of different families that may be present and the complexity of the structure within each family.

Bio-Linux is an integrated, bioinformatics-centred, research platform. By providing both standard favourite and cutting edge bioinformatics tools on a Linux-based system, it combines the benefits of being powerful, configurable, and easily updateable, with the ease of use and potential for software integration required for the handling and analysis of biological data.

Radiation hybrid maps coupled with the mapping of expressed sequence tags and their organization into UniGene clusters, has revolutionized the way comparative maps are built and maintained. We have used publicly available rat, mouse, human, and zebrafish data to build completely integrated comparative maps.

The GET3D, Genomic Exploration Tool in 3D, software tool is a complimentary product to the CAS, Categorised Annotation Set, system. Through 3D visualization and interaction techniques it allows for the manipulate of the CAS dataset, including the highlighting and exploration of data relationships, and for adding new information and relationships.

Quantitative real-time PCR represents a highly sensitive and powerful technique for the high-throughput analysis of virus load and gene expression. We used the Q-Gene software (Muller et al., 2002) to expedite the statistical analysis, graphical presentation and evaluation of the real-time PCR quantitation of Nipah virus in experimental animals.

TAPiR is a novel algorithm for identifying, clustering and visualizing time-course microarray experiments based on either fold change or t-test p-value thresholds. TAPiR assigns letters to specific patterns of change that can be oriented along the time-course to form “words” that can then be sorted alphabetically to identify similar clusters.

We apply correlation analysis, a tool developed for analysis of functional magnetic resonance images to microarray timecourse experiments. Although fMRI and microarray timecourses share little in common physically, the data analysis tasks are quite similar, and correlation analysis is shown to be a useful preprocessing tool to apply prior to clustering gene expression timecourses.

The Knowledge Discovery Tool or KDT is a tool that utilizes a knowledge modeling approach to intelligently extract and integrate a large set of disconnected bio-medical information. Its unique user interface allows it users to see how the entities are linked together, uncovers hidden linkages and highlights certain unusual links that might be worth exploring.

The sequence identifier end set cardinal is proposed as a new estimator of linear sequence linguistic complexity. A scanning window algorithm to compute this value is presented and profiles obtained with genomic sequences are discussed.

SNPview, the interactive navigator for the individual SNPs, SSLPs, alleles and haplotypes projected to the genomic axis is available on-line at http://www.gnf.org/SNP/ . Large, but discrete regions of the genome are not very polymorphic between particular strain pairs, and thus cannot easily be interrogated for natural genetic variations influencing QTLs.

We developed a computer aided molecular modeling system using virtual reality technologies. Although it is still a prototype, the most characteristic function of the system is enabling its user to “touch” and “feel” the electrostatic potential field of a protein or a drug molecule.

BIRCH is a  resource of integrated programs and databases for molecular biology, unified through the GDE graphic interface. The BIRCH framework is designed for semi-automated installation and customization on Unix systems, and integration of locally-installed software and databases into BIRCH. http://home.cc.umanitoba.ca/~psgendb

SVG has enabled a visually appealing application for the visual comparison of Brassica napus and Arabidopsis thaliana. SVG overcomes two key drawbacks of current web-based genome browsers: fixed displays and frequent page reloads. The Brassica Araboidopsis Comparative Genome Browser is available online at http://www.brassica.ca.

STING Millennium (SMS) and Java Protein Dossier (JPD) make a powerful duo for the structural analysis of macromolecules. SMS is a web based set of programs and databases for analysis of protein structures, while JPD provides an abundant collection of physical and chemical descriptors/parameters. SMS/JPD v.3 is available at http://www.cbi.cnptia.embrapa.br

We developed a framework to visualize SNPs, haplotype blocks, and genes across chromosomal physical maps and their relationship with linkage disequilibrium maps. This visualization is aimed to the cost-effective selection of SNP markers for disease association studies as a function of the profile of LD obtained on reference population samples.

It was shown from Inverse Flux Analysis (IFA) of Escherichia coli that if pyk was knocked out, then the flux through ppc and TCA cycle pathway increased, and the acetate production flux reduced but acetate production increased while ppc was knocked out. The experimental data well coincided with the IFA results.

I. We have developed a system that monitors various data sources, dynamically extracts gene information, comprehensively matches genes, and integrates them into a central database by categories, such as pathway, genetic mapping, phenotype, expression profile, domain structure, protein interaction, disease association, and references. The system achieves high performance when querying a large batch of genes together

ELM is a resource for predicting functional sites in eukarytic proteins. Putative functional sites are identified by conventional methods, such as patterns (regular expressions) or hidden Markov models. To improve the predictive power, context-based rules and logical filters will be developed and applied to reduce the amount of false positives.

FlyMine is a project to build an integrated database of genomic, expression and protein data for Drosophila and Anopheles. Data are stored massively redundantly and arbitrary queries are allowed using a web interface or Java API. Queries are re-written in real time to make use of redundant tables.

InterPro is an integrated protein signature resource for predicting protein families, domains and functional sites. It incorporates data from PROSITE, Pfam, PRINTS, ProDom, SMART, TIGRFAMs, PIR Superfamilies and the structure-based SUPERFAMILY. InterPro classifies 80% of SWISS-PROT/TREMBL and provides links to the proteins, methods, specialised protein family resources and structural information.

The dbZach System is a microarray information management system meant for local installation in labs or larger, enterprise environments. The dbZach database includes four core subsystems, and six modular subsystems. The system also includes GUI data mining tools, and an API. Source code will be available soon at: http://dbzach.fst.msu.edu.

MyMED is an internal relational XML database implementation of MEDLINE citations. The MyMED database is necessary to execute text mining algorithms and complex text searches in a fast, secure manner. Data is stored in a DB2 database that is enabled for the XML Extender and Text Information Extender.

The Rat Genome Database (RGD) provides research groups access to rat genomic and genetic data, including annotated sequence, related to particular diseases. DORR will help RGD answer the need of the rat community for access to curated data of interest. rgd.mcw.edu

The Rat Genome Database (RGD) is a disease centric resource of comprehensively curated data from rat genetic and genomic research. RGD's goal is to provide information that will aid researchers in using the rat as a model organism for human disease studies. RGD is available at rgd.mcw.edu .

Webin, Webin-Align and SPIN are the web-based tools for submitting nucleotide sequences, nucleotide sequence alignments and protein sequences respectively to EMBL-Bank, EMBL-Align, or SWISS-PROT databases. These tools guide you through a sequence of WWW forms allowing interactive submission. All the information required to create a database entry is collected during this process. All submission tools are available at http://www.ebi.ac.uk/Submissions/index.html

The UniProt Consortium (European Bioinformatics Institute, Swiss Institute of Bioinformatics, and the Protein Information Resource) was created to merge Swiss-Prot, TrEMBL and PIR database activities into UniProt, a comprehensive resource of protein sequences and function. UniProt has three layers: protein sequence archive, protein knowledgebase, and non-redundant reference (NREF) databases.

PRIME(http://prime.ontology.ims.u-tokyo.ac.jp/) is an integrated database involving major completely sequenced eukaryotes. It contains the protein-protein/gene/compound interaction data extracted by natural language processing, domain information, structural information, protein kinase classification, and ortholog tables among organisms. The comparison and prediction of pathways are also available by an automatic pathway graphic image interface.

CAS is a data integration system that allows for the creation of categories and integrated data sets. Each category consists of a set of attributes and methods, which act as an object item. Incoming data, obtained manually or through the automated data collection component, is linked to category attributes according to user defined rules and conditions.

We propose protein structure modeling using a geographic model and build a structure database which includes thematic information and geometry of protein. In the modeling, geometry is represented by spatial types and thematic information includes the physico-chemical data. We state queries to retrieve topological relationship between structural elements with spatial operators.

We propose modeling of sequence versions for sequence changes of the same piece of DNA using a time stamp attribute in a temporal data model and mechanism of management of sequence versions in a sequence database by applying trigger rules(Event-Condition-Action) in an active database system.

GCC is a database for immune cells transcriptomes. It is based on a database system (built using opensource technologies) that allows for data storage and intelligent retrieval. It is targeted at the affymetrix platform and allows for MIAME compliant experimental annotation and supports adoption of ontologies.

dbSTR is a repository of short tandem repeats (microsatellites) whose polymorphisms have either been predicted using machine learning or verified using wet-lab methods. These STRs could be useful markers for high resolution linkage and association studies. dbSTR is freely available at http://www.dbstr.org.

BIOSILICO is a web-based database system that facilitates the search and analysis of metabolic pathways. BIOSILICO allows efficient retrieval of all available information on enzymes, compounds, reactions and pathways by integrating the heterogeneous metabolic databases and generates well-designed view pages showing retrieved data in a systematic way for easy understanding.

GlycoSuiteDB is a relational database of published glycan structures designed to assist researchers in the analysis of glycans. GlycoSuiteDB can be accessed from http://www.glycosuite.com

SNP PrimerPicker is a software system for designing primers. Sequences in record are aligned with the source using bl2seq. It utilizes Primer3 to find primers and blastcl3 to compare similarity with the chromosomes. The components are integrated to make the procedure convenient. It is web accessible at http://bcz099.ust.hk/primerpicker/.

The Melbourne Brain Genome Project is an Internet resource for studying gene expression as measured by serial analysis of gene expression, in both normal mice and specific mouse models. These models mimic neurodegenerative human diseases. The resource includes tools developed to analyse this data and is available at http://www.mbgproject.org.

We describe a tool to manage assembly processes. Two main advantages of this system are: [i] Based on "assembly projects" which contain all the relevant data of an assembly. This allows for re-assembly at a later point with (e.g.) additional input sequences. [ii] All user interaction is through a GUI.

Based on the composing of php, Mysql, Linux, we construct the S. cerevisiae (15,000 entries) and putative C. elegans protein interactions database. This database can help to annotate novel proteins by the interacting partners; also can provide proper candidates to narrow down the scale of further high-throughput screening experiments.

The EBI provides free, publicly available bioinformatics services for the scientific community. These can be divided up into the following categories: data submissions processing, biological database production, access to query, analysis and retrieval systems, ftp downloads, training and education and user support. These services are available at: http://www.ebi.ac.uk/services.

KEGG API is a new web service for accessing the KEGG database. Using the APIs in the local program, the user can retrieve various information about genes, pathways, chemical compounds etc. stored in the latest versions of the KEGG database. KEGG API is available at http://www.genome.ad.jp/kegg/soap/.

We present a novel approach to the integration of distributed molecular biology information resources, which consists in a design of an adaptive natural language interface and application of multiagent technology. Our approach permits integration of any databases which have a common subject domain. The implemented prototype is available at http://urchin.spbcas.ru/NLP/NLP.htm.

We present a module (JPIPE) that allows EMBOSS users to build analysis pipelines under the JEMBOSS GUI. JPIPE is a flexible workflow system that complements JEMBOSS. A tracking system is a part of JPIPE so the user is able to recreate pipelines, compare results at a particular point of the workflow and administer ongoing jobs.

pBIRD is a fully-searchable web based system designed to identify, catalog and describe a broad range of available bioinformatics software. The system is designed to upload, store, track and retrieve information associated with and about bioinformatics software products deposited in the database from both internal and external sources.

As molecular genotyping technologies accelerate there is an increasing need to communicate precise information on polymorphism data in machine-readable format. We have developed a hierarchical object model and XML implementation called Biological Variation Markup Language (BVML) to facilitate exchange between genotyping laboratories and distributed databases.

We developed the updateBASE system which provides real-time, automatic updating of biological databases under the client-server architecture. Using this system, an annotator can get up-to-date database sources and get more confident annotation results. The updateBASE will play an effective supporting role in predicting functions and relationships of unknown sequences.

CleanBank is a database that documents suspected artifacts found in sequences and/or their annotation in the international sequence databases. The artifacts are either reported by researchers, or identified by curated algorithms. Current algorithms detect E. coli and vector contamination. For a detailed description and a preview, see http://bip.weizmann.ac.il/MIW/CleanBank/index.html

We present a solution to the huge variance problem that has plagued B-tree supported genome interval overlap queries. Our approach is based on translating the overlap query into a two-dimensional point-in-box geometric query supported by an R-tree index.

A database that formalizes Signal Transduction Pathway knowledge in scientific literatures is presented. The database focuses on annotating pathways or sub-pathways according to their related biological processes. Every process and element in a pathway has a pointer to ontologies, such as GO, and one can search (sub-)pathways, molecules by using them.

ANTIMIC is a specialized database dedicated to antimicrobial peptides. It contains useful analysis tools that can aid wet-lab scientists to determine the family and function of a putative new anti-microbial peptide and also in design of artificial anti-microbial peptides.

We report the construction of a novel database termed OrthoDisease, which was constructed using the Inparanoid program to analyze a list of disease genes derived from the Mendelian Inheritance in Man database. Our database is accessible online at orthodisease.cgb.ki.se and can be searched according to disease/gene/protein name or EC/MIM number.

BASE is a free microarray database system with a clean and intuitive web interface. It manages biomaterials, array production and raw data with images. Analysis tools are included, and users can provide new tools through a plugin interface. The BASE web site is http://base.thep.lu.se/.

MARS is an intelligent diagnosis system for human genetic disease. The MARS consists of databases of human genetic disease information and mutation detection system. The first release of MARS contains genetic information of MECP2 gene and its website is available at http://www.genome.re.kr/mars/

SDPS database is a comprehensive structural database of small disulphide-bonded proteins. This database is enriched with a number of new features which cannot be easily accessed through public databases. The database aims to facilitate the research on small disulphide-bonded proteins especially on disulphide connectivity features. SDPS database can be accessed freely at http://origin.bic.nus.edu.sg/sdps.

BioPAX (http://biopax.org) is a new community-based initiative to address the growing need for a unified framework for sharing pathway information.  Several groups are participating in BioPAX to develop a data exchange format that will allow communication between existing pathway databases and facilitate deposition of data into a common public repository.

The Mouse Genome Informatics (MGI) databases provide access to comprehensive, integrated, experimental data for the laboratory mouse in the domains of sequence, expression, gene function (GO), molecular variation, phenotype, inbred strain characterization, homology, and tumor biology. MGI provides the definitive mouse gene index. MGI can be accessed at http://www.informatics.jax.org/.

A measurement with edit distance is a typical way for searching for similar DNA sequences. But in some cases, time warping distance is more appropriate for measuring similarity between two DNA sequences. In this paper we propose a query method based on time warping distance by coding DNA sequence.

This research will investigate the ways in which biologists analyse data in their plant studies, and how these requirements may be expressed through a visual query language that allows researchers to directly address the plant characteristics that are of interest.

Data warehousing using dimensional modeling and s tar schema data marts was applied to biological information. The data marts capture biological relations hips like gene--protein, as well as numerical data in the intersection between dimensions for example expression data.

Focus was also on building a automatized extraction and transformation of a wide range of public data sources.

SeqHound is a resource containing daily updated Entrez databases and 3-D structural data. It holds links to similar sequences, taxonomy, complete genomes, functional annotation, structural domains and literature. SeqHound is accessible directly through a C API and via a web server through PERL, Bioperl, C or C++ APIs.

The GENA Genomics Array database stores data generated by the Microarray process as well as information describing the experimental conditions. GENA provides structured queries to extract meaningful data that supports comparative analysis of gene expression ratios. Gena is unique in its capacity to mine gene expression data from several perspectives.

The goal of the ScriptSure project is to create a database that gives an accurate, comprehensive representation of the human transcriptome. ScriptSure provides a non-redundant representation of the transcript data, provides high quality (genomic) sequence, and alleviates problems associated with current approaches to representing transcript data. http:://sapiens.wustl.edu/ScriptSure

The Unified Modeling Language (UML) can be used to display Bioinformatic data objects and their relationships graphically. The first step in the process is done at the conceptual level, allowing domain experts like biologists to participate. In this paper, we sketch the process of creating an XML document from scratch.

We have built a first-generation computational tool for siRNA selection (http://jura.wi.mit.edu/bioc/siRNA) which implements sophisticated selection algorithms to identify siRNAs with a high probability of specifically silencing the target gene. We have also designed a prototype database called sirBank to be a repository for siRNA molecules known to silence target genes.

We found strong evidence (p<0.01) of cancer-specific splice variants in 316 human genes through a genome-wide analysis of human expressed sequences. The majority of these genes have functions associated with cancer. For a large number of cancer-associated genes, it appears the normal form instead of the cancer form that is previously uncharacterized.

Identification Of Novel Two-partner Secretion Family in Burkholderia pseudomallei Filamentous hemagglutinin belongs to the Two Partner Secretion (TPS) family. Sequence analysis indicated the presence of filamentous hemagglutinin (FhaB) and its transporter (FhaC) in an operon in Burkholderia pseudomallei. Motif recognition, phylogenetic analysis using N-J and PAM matrix methods provide more information of the functionality of the operon.

We describe our concept for the integration of heterogeneous data into a platform for systems biology. We have implemented a Bioinformatics Resource for the Integration of heterogeneous Data from Genomic Explorations (BRIDGE) and illustrate the useability of our approach as a platform for systems biology for two sample applications.

We propose an algorithm, which reconstructs gene networks from expression data, trying to face the problem of "small" available data, assuming some reasonable biologically consistent hypotheses. We also evaluate algorithm performance on artificial problems.

We present a procedure that integrates knowledge of multiple biological databases to recover potentially involved pathways from time course microarray experiments. Starting with a refined new clustering algorithm we group similarly behaving genes, followed by an integrated analysis of common transcription factor binding patterns, functional categories and biologically verified pathways.

We present an algorithm for combining genome-wide expression and protein-DNA binding data to discover co-regulated modules of genes and associated regulatory networks. Our algorithm operates on discovered networks to label transcription factors as activators or repressors, identify patterns of combinatorial regulation, and uncover sub-networks for biological processes in Saccharomyces cervisiae.

We analyzed conservation of alternative splicing patterns in pairs of orthologous genes from the human and mouse genomes. Our results demonstrate considerable diversity of alternative splicing in these genomes: at least half of alternatively spliced genes have species-specific isoforms. Orthologs with non-conserved isoforms may play a role in species-specific development.

To cope with the need for automated data conversion, storage, and analysis in the field of proteomics, the open source system ProDB was developed. The system handles data conversion from different mass spectrometer software, automates data analysis, and will allow the annotation of MS spectra.

In functional genomics, many are worried at the number of functionally unknown genes we have. One of the areas that we think has contributed to this is the uncertainty in low sequence alignments (twilight zone). Our work involves the development of a rule-based system that allows us to accurately assign functional annotation in the twilight zone.

We took a probabilistic decision tree approach to predict co-complexed pairs (CCPs) of proteins by integrating high-throughput interaction datasets with other characteristics of gene/protein pairs. Our method made more sensitive and specific predictions than high-throughput interaction screens, and is also promising in detecting unknown CCPs.

To improve protein function prediction from high-throughput, error-prone data, we compute a posterior probability for the validity of each interaction. These edge weights are based on the experimental data and how well each observation fits the expected network topology. Our function predictions compare favourably to previously published methods.

We present GOPArcII, a new version of our comprehensive, open source framework for the integration of functional classifications and metabolic pathways. GOPArcII is based on a relational database. It enables a user to search and handle data like genome data from the perspective of functional categories and metabolic pathways.

The computational recognition of precise splice junctions is a challenge faced in the analysis of newly sequenced genomes. To understand the sequence signatures at the splice junctions, comparative analysis using both neural network based calliper randomization and information theoretic based feature selection approaches have been used.

LOC3D is both a database and a web server for predicting the sub-cellular localization of eukaryotic proteins of known structure. Localization is predicted using a combination of four different methods; prediction of nuclear localization signals, through sequence homology to proteins with known localization, automatic text analysis of SWISS-PROT keywords and using neural networks.

Overlapping 9 and 15mer peptides covering the insulin receptor alpha-subunit sequence were synthesised and measured for their ability to specifically bind ¹²⁵I-insulin. The insulin binding sequences were analysed to identify putative insulin-binding regions of the receptor, insulin-binding motifs, and develop a preliminary insulin-binding scoring matrix.

Annotation of novel genes is a challenge to genome sequencing efforts. We streamlined the process for genes with novel protein-coding domains by integrating web-based databases and annotation tools with gene data from diverse sources. We show examples of how our approach enhances experimental design and leads to accurate gene annotation.

It is likely that multiple genetic variants interact to confer susceptibility to complex disease. We have shown that functional variants in the MTHFR and ACE genes interact to increase risk of migraine.

We constructed a system that lists candidates of readthrough genes based on the existence of a “protein motif” at the 3’UTR. Using this system, we extracted 85 candidates in Drosophila melanogaster, and found features in those sequences which are known to have an effect on readthrough events.

GOODIES is a Gene Ontology-based data-mining tool for effective functional classification. Given biological entities, GOODIES classifies them along their annotational attributes and selects optimal GO candidate terms from combinatorially many choices for overall biological interpretation, with intuitive visualization. The major applications of GOODIES include biologically-oriented cluster analysis and functional categorization.

Phylogenetic profile analysis assigns functional clues to proteins in a manner that is independent of sequence similarity. We present an improved algorithm that constructs the phylogenetic profiles of proteins based on the unambiguous Smith-Waterman Alignment algorithm. We used MetaCyc metabolic pathway clusters to estimate the prediction accuracy of the method.

It is widely accepted that our complexity as a species results from the regulation of our genes. Our approach is to analyse, on a genome-wide scale the upstream regions of human genes with particular emphasis on transcription factor binding sites. The ultimate goal is to infer expression pattern from sequence.

Adjacent human gene co-expression was investigated using GeneLogic library. We analyzed average and maximum expression profile correlation for gene pairs within sliding genomic windows for normal or diseased tissue samples. Significance estimates used randomized gene sets. Co-expressed clusters were analyzed for gene duplication, subcellular localization, functional themes, inter-species conservation.

Lactobacillus plantarum is a versatile lactic acid bacterium that is encountered in various niches. LacplantCyc is a newly created pathway/genome database predicted from the annotated complete genome sequence of L. plantarum WCFS1 (PNAS 2003;100:1990), using the PathoLogic software from PathwayTools (Peter Karp). Manual editing and experimental verification is in progress.

The use of Boolean models is explored in reconstruction of the topology of genetic transcriptional networks, employing gene expression data from simulated perturbations. The construction and employment of a software suite for such exploration is described. Results are compared for different simulated topologies, inference methods, and amount of noise.

We would like to report a scalable multi-processor application for clustering of gene expression profiles. The program implements the flexible clustering algorithms developed at PBRC. The program is written in MPI standard and tested on IBM AIX Parallel Environment.

We wish to detect signals in mRNAs that influence their translation. These include signals that modulate translation efficiency, mRNA stability or its localisation. To do this we have developed the TransTerm database (http://transterm.otago.ac.nz/) of mRNA regions and regulatory elements. We have detected novel elements and are testing them in vivo.

PATIKA is an ongoing research and development project for collaborative construction and analysis of cellular pathways. Our software tool provides an integrated, multi-user environment for visualizing and manipulating network of cellular events. PATIKA is available at http://www.patika.org.

We obtained 3’ end partial sequence of cDNA which have not found homologs in ‘nr’ by using BLAST. We predicted full length genes from partial sequence and cloned full length genes by using a predicted sequence

Antisense oligonucelotides are an important tool for gene-knockdown approaches in functional genomics. We assess the usefulness of current approaches predicting accessibility and/or efficacy using a database of known results. A set of utilities developed for AO and siRNA design (control design, specificity, site selection) is available at http://sonnhammer.cgb.ki.se.

We devised an automated computational procedure to detect pairs of overlapping and oppositely directed human transcriptional units (sense-antisense pairs) equivalent to a large set of putative mouse sense-antisense pairs previously predicted by cDNA mapping. We found support in the human transcriptome for a significant proportion of mouse sense-antisense pairs.

We have developed an integrated software platform that addresses data generation and handling, data verification/quality control and bioinformatic analysis steps of large scale proteomics projects. Experimental data is integrated with computationally enriched data from public and proprietary databases enabling protein isoform distinction, protein-protein interactions analysis, pathway analysis and text mining.

LION Target Engine is a user-friendly system designed to streamline the identification and validation of targets in an enterprise environment. The system components for: sequence registration and curation; sequence analysis; gene and protein index; text mining; pathways and interaction networks; 3D structures; TaqMan and microarray data handling, and target tracking.

Small interfering RNA (siRNA) is used in functional genomics applications to produce “knock-down” cells. The siRNA design tool scans a target gene for candidate siRNA sequences that satisfy user-adjustable rules. Selected candidates are then screened to identify those siRNA sequences that match only the gene of interest.

T-DNA use in plant functional genomics is based on a hypothesis that the T-DNA randomly inserts plant genomes. To test this, we analyzed 120,000 T-DNA flanking sequences of the SIGnAl database. Of the total 29,084 Arabidopsis genes, approximately 70% have >1insert, whereas 8760 (30%) genes are still left without any.

A network analysis of the protein-protein interactions in yeast reveals distinct clusters. Some of the clusters are due to functional groups, while most are due to methods of detection; interactions detected by yeast two-hybrid tend to cluster proteins into groups that are different from the clusters due to mass spectrometry. We quantify these differences and discuss implications.

cSAGE is an open-source application written in C to provide an efficient mechanism for extracting SAGE tags and assigning matches to DNA sequences. It has been used for a cold tolerance experiment in A. thaliana involving 3 librarys with more than 180,000 tags. See http://homepage.usask.ca/ctl271/csage for more information.

We describe a computational approach to the reconstruction of the architecture of a rearranged genome based on data from end sequence profiling experiments. We apply our techniques to the reconstruction of the genome of a human MCF7 tumor cell.

We describe the identification and classification of the complement of protein kinases and phosphatases in mouse. We also present preliminary results from a functional screen of these proteins, coupling sequence based classification with high throughput functional screens.

BioinformatIQ® is an integrated system for handling the information needs of proteomics from sample preparation, through automation of instrumentation, to protein identification and characterization. This informatcs platform for proteomics is demonstrated through application to the proteomic analysis of human plasma. More information on BioinformatIQ® can be found at http://www.proteomesystems.com

YETI is a novel bioinformatics tool for integrated visualization and analysis of functional genomic data from the yeast Saccharomyces cerevisiae. YETI 1.0 consists of three fully inter-linked sections allowing users to explore the “genomic” (e.g. chromosomal location), and “proteomic” (e.g. associated protein-protein interactions) context of multiple proteins of interest simultaneously.

Three approaches for the in silico prediction of UTR repeats have been used on a test data set, resulting in the detection of sequence stretches in ~5% of the input sequences during clustering and reduction in size of large clusters. Seven of those putative repeats have been proven to be repetitive in vivo by Southern blot analysis.

We present a hypothesis-independent method for identifying the region of a genome coding for a protein sequence using proteomic information. The method can be used to identify novel genes that were not found by other annotation techniques. It is demonstrated using theoretical and experimental data sets from prokaryotic and eukaryotic organsims.

We have developed the new interaction generality measure (IG2), which can be used to computationally assess the reliability of the interactome data. We performed an integrated analysis by using comprehensive phenotype dataset and IG2-treated interactome dataset from yeast, which yielded global insights into the biological features of the protein complexes.

Using the new criteria for CpG islands introduced by Takai and Jones, we investigated several association types between CpG islands and genes to further establish the importance of CpG islands as gene markers. Our investigation gave us a useful tool for evaluating the accuracy of gene annotation in human chromosomes.

cDNA2Genome is a web tool for automatic high-throughput mapping and characterization of cDNAs. It uses already existing annotation data and improves them when possible in the case of ESTs, proteins and mRNAs. It is focussed on the determination of the cDNA exon-intron structure. The final result of cDNA2Genome is an XML file with all information obtained by the task.

Disease related Genomic Regions Of Interest (GROI) were submitted to the Rat Genome Sequencing Consortium for prioritized sequencing. These areas were analyzed to determine whether greater coverage facilitated denser and more accurate annotation. Functional annotation of genes in these regions was achieved using datamining of public databases and manual curation.

DNPTrapper is a graphical tool specifically designed for finishing shotgun assemblies containing complex repeated regions. DNPTrapper allows visualization of sequences and sequence features, e.g DNPs, mate-pairs, repeat boundaries, chromatograms, etc. in horizontal and vertical representation, followed by manual and semi-automatic manipulation, which greatly simplifies finishing.

The e-Protein project aims to provide structure-based annotations of proteins in the major genomes by linking resources via Grid technology at 3 sites; Imperial College London, University College London, and the EBI. At the end of the first 6 months we have established a pre-prototype using GLOBUS/Grid technology.

Gene Ontology Toolkit is a Java GUI for working with the Gene Ontology (GO). Its three integrated components (editor, annotator and merger) enable biologists to visualize and extend GO, annotate gene products with GO terms, and merge the extensions and/or the annotations with new release of GO or others’ extensions.

We have completed the assembly of the Leptospira borgpetersenii serovar hardjobovis genome, which contains a large chromosome (CI) of 3,614,529 base pairs and a smaller chromosome (CII) of 317,585 base pairs. The annotation of CII is complete and is currently in progress for CI.

We have implemented the Ensembl genome sequence database and interface for handling the annotation of the bovine genome. Initial efforts have focussed on display of the annotation of small and micro RNAs and linking to the SNP database, IBISS.

To identify members of known families of RNAs a combination of BLAST and INFERNAL is used with the RNA covariance models in Rfam. To identify potential new RNAs a combination of genome specific BLAST and QRNA is used. The results of analysis of bovine BAC-end sequences will be shown.

ASmodeler is a novel web-based utility to find gene models of alternative splicing events from genomic alignment of mRNA and ESTs. It can be used as a transcript assembly program, an EST clustering utility, and a method of comparative gene modeling. ASmodeler is available at http://genome.ewha.ac.kr/ASmodeler/.

We present a fast, fully automatic procedure for protein domain extraction from single query sequences without pre-calculation of domain statistics. Combinatorial clustering of domains from BLAST hits using quasi-convex set functions, followed by domain parsing and pattern discovery permits highly efficient (polynomial time) whole genome functional annotation.

Our web-based genome annotation system has major modules such as gene prediction, homology search, promoter analysis, motif analysis, gene ontology analysis, annotation databases, and a genome browser which shows the entire information of a genome. We have also developed a motif analysis and a gene prediction programs based on HMM.

We propose a new method for increasing the gene prediction accuracy without using information of known genes. To increase the gene prediction accuracy we used the additional learning data through the phylogenetic concept. Tests on 3 complete prokaryotic genomes performed with the GLIMMER program demonstrate the ability of the new approach to detect additional genes.

A general and flexible multi-motif model is proposed for promoter motif analysis based on dynamic programming. By extending the Gibbs sampler to the dynamic programming and introducing temperature, an efficient algorithm is developed for searching motifs in promoters. The algorithm is tested with plant promoters.

The aim of this study is a development of a suitable procedure to predict the function of viral genes. To overcome conventional searches based on similarity, HHV (Human Herpesvirus genomes) were analyzed by COG and phylogenomic methods. It will provide a practical method to predict the function of new genes in viral genome.

The unified vocabulary of terms provided by the Gene Ontology consortium has become a standard tool in annotation of genes and their products. We present a web-service available at http://goblet.molgen.mpg.de, which allows annotation of anonymous cDNA or protein sequences by GO terms.

A bovine in Silico SNP database has been constructed. Contigs of ‘unique bovine sequences’ were established which were treated as model mRNAs. A comprehensive web interface has been developed which highlights putative identity of each contig, putative SNPs, location of predicted intron-exon boundaries, and genome mapping data for each model mRNA.

The Encyclopedia of Life Project(EOL; http://www.eolproject.info) is aimed at utilizing Grid computing resources to catalog the complete proteome of every living species in a flexible, powerful reference system using a scalable protein annotation pipeline. Recognized protein sequences are assigned putative functional annotation, structure assignment, and cross-referenced to other data sources.

A new tool called System for Automated Bacterial Integrated Annotation - SABIA was developed for the assembly and annotation of bacterial genomes. This system performs automatic tasks of assembly analysis, ORFs identification/analysis, and extragenic regions analysis. Genome assembly and contigs automatic annotation data are also available in the same working environment.

The patterns of transcription factor binding sites (TFBSs) within upstream regions of differentially expressed genes identified via microarray analysis may help identify regulatory genetic networks. We have thus created a database of putative TFBSs conserved across orthologous human, mouse and rat genes through the use of MAVID, Match and Transfac.

We present a fully-automated procedure to create a protein structure family database, along with a corresponding homology searching algorithm based on a combined E-value/Z-score approach. Whole-genome struture-based annotation tests on several completely sequenced genomes have demonstrated results comparable to PSI-BLAST.

The Disease Ontology is a hierarchical controlled vocabulary created to represent human disease. The ontology was created in order to enable database curation of disease gene associations. All terms in the ontology also maps to billing codes for the purpose of medical record mining. The ontology is available at: http://sourceforge.net/projects/diseaseontology/.

To generate meaningful mRNA expression ratios, data from separate arrays or probes must be normalized. Median normalization performs adequately only when differences between data sets are linear. To cope with non-linearities and noisy data, I have implemented a binned-intensity normalization algorithm which outperforms simple median normalization.

The robust clustering of some normal samples within tumor groups and robust clustering of other normal samples in a separate, 'normal' group indicates the confounding of control samples. Our approach uses the maximum difference subset algorithm (MDSS) and bootstrap validation, which evaluates the difference in mean expression between two groups.

We propose to structure analysis of microarry according to biological knowledge in order to provide an intuitive and biologically meaningful rationale for computational classification results. Thereby, we rely on the Gene Ontology to attribute genes to biological aspects and usual machine learning methods for classification.

We validated a quenching method for the harvesting of micro-organisms from liquid cultures for gene expression studies. The transcription profiles of quenched L. plantarum WCFS1 cells were compared with those of cells that were harvested by alternative methods. PCA analysis and hierarchal clustering of the resulting transcriptomics data show a clear effect of this quenching method on the transcription profiles.

We have developed a method for predicting transcription factor responsive genes by combining response element prediction with cDNA microarray data. Our Statistical Promoter REgulatory Element (SPREE) application program identified cAMP responsive elements. SPREE output was combined with cDNA microarray data to design an SVM model for predicting cAMP responsive genes.

RAD and its web-based annotation forms are a system aimed at the collection, organization, and exchange of all relevant information pertaining to gene expression array (and SAGE) studies. The richness of information captured and the use of ontologies render RAD a very powerful infrastructure for querying and analysis ( http://www.cbil.upenn.edu/RAD3).

MPrime is a system for efficiently creating large sets of PCR primer pairs for use in designing products for custom cDNA microarrays. MPrime has allowed us to effectively design custom neurodegenerative microarray chips for humans as well as the rat and mouse genomes. MPrime is available at: http://kbrin.a-bldg.louisville.edu/Tools/MPrime/

We train a Bayesian network to represent statistical dependencies between gene-expression levels from DNA-microarray datasets. The trained network is used to predict the effect of pathologically altered expression levels on the global expression pattern (inverse modeling). By use of this ability we can powerfully predict new genes involved in pathogenesis.

We present a method for estimating key regulatory genes of genetic networks by analyzing their network topology. In networks learned from childhood leukemia microarray datasets we find a small number of genes such as POU2AF1 that may contribute to B-cell tumorigenesis.

We used a Bayesian Network learning technique to analyze microarray transcriptional response profiles of yeast to nitrogen oxide. Using gene clusters as network nodes, the learned transcription response networks are consistent with the proposed biological hypotheses. The model also revealed a previously unknown link between galactose input and a fzf1 dependent cluster.

GEPAS is a web-based pipeline for microarray gene expression profile analysis, freely available at http://gepas.bioinfo.cnio.es. The most commonly used tools for the processing and management of different functional genomics data are included in GEPAS as interconnected modules that exchange information in a user friendly manner.

The existing k-means clustering algorithm for gene expression data suffers from its uncertainty and ambiguity. This robust k-means clustering algorithm demonstrates how both variations in data sources and the intrinsic indeterminacy of clustering procedures can be overcome and that reliable, informative, and optimal clustering results can be achieved.

Selection of spotting-buffer and slide-chemistry is critical for the reliability of microarray-hybridization-experiments. Comparisons, however, have tended to be subjective, not suitable for systematic study. We present a novel approach, objectively assessing spot-morphology and -variance by measuring average (variance) of radial spot-pixel-intensity-profiles. It is successfully demonstrated comparing over 24x6 buffer/slide combinations.

We present a probabilistic method for identifying regulatory modules from gene expression data. Our procedure identifies modules of coregulated genes, their regulators and the conditions under which regulation occurs, generating testable hypotheses in the form ‘regulator X regulates module Y under conditions W ’. We present microarray experiments supporting three novel predictions,suggesting regulatory roles for previously uncharacterized proteins.

The use of different terminologies and structures in microarray databases is limiting the sharing of data and the collating of results between laboratories. We have proposed an integrated information management architecture for microarray experimental data that will focus on addressing these problems.

This concise guide demonstrates the implementation of BASE as a microarray data analysis pipeline. BioArray Software Environment is an open-source platform for archiving, analysis and visualizing of microarray data. Data for this demonstration compares gene expression between a wild-type and mutant Arabidopsis plants that identified 86 differentially expressed genes.

We propose a statistical method for estimating a gene network based on Bayesian networks from microarray data together with biological knowledge including protein-protein interactions, protein-DNA interactions, binding site information, existing literature and so on. Our method can optimize the balance between microarray and biological knowledge automatically.

PaGE is an algorithm we have developed at CBIL which uses statistical methods to assign discrete patterns to gene expression data. This poster will highlight the new implementation (version 5.0) with improved novel FDR alrogithm based on the Westfall and Young minP stepdown distributions, and new interface.

GenMAPP is designed for viewing expression data on biological pathways. GenMAPP automatically color-codes the genes according to criteria supplied by the user. MAPPFinder matches expression data to the Gene Ontology and indicates whether there is a significant over-representation of genes meeting the user’s criterion for each GO term. http://www.GenMAPP.org.

A decision-tree learning approach was applied to classify three types of prostate tumor tissue samples using microarray gene-expression data. In LOOCV, results were comparable to those obtained from applying SVMs or weighted voting method to this dataset. Furthermore, human-understandable models from decision-tree learning correctly predicted sample classes in previously published prostate tumor datasets.

The hox gene HB24 encodes a protein expressed in CD34 cells which plays a role in T-cell activation. Using a statistical approach based on regression analysis we identified 29 genes co-regulated with the HB24 gene. The analysis of the regulatory regions of these genes revealed common transcription factor binding sites.

BioRag (Bio Resource for Array Genes at http://www.biorag.org) is an interactive platform for analyzing and developing a biological interpretation of the microarray results. Differential gene expression patterns can be interpreted using tools that mine and extract variety of biological relationships captured in this integrative resource.

We present QUINTET: an R-based unified cDNA microarray data analysis system with graphical user interface. It can seamlessly perform five principal categories of the data analysis: data quality assessment, faulty spot filtering and normalization, identification of differentially expressed genes, clustering of gene expression profiles and classification of samples.

We developed a statistical method for estimating gene networks and detecting promoter elements simultaneously. The estimation of gene network from cDNA microarray data alone is likely to cause misdirected edges. Our method overcomes this problem by integrating microarray data and the DNA sequence information into a Bayesian network estimation.

To study the transcriptome of the malaria parasite, we designed a custom no-mismatch oligonucleotide array containing 260,596 25mer single stranded probes from predicted coding sequence (including mitochondrion and plastid genome sequences) and 106,630 probes from non-coding sequence. In addition 124,957 probes from Plasmodium yoelli contigs are include on the array.

This shows improvements on the best current estimates for gene abundance using Affymetrix raw data, by compensating for spatial heterogeneity, and by assessing individual probe quality and background, prior to using a robust method for fit.

It is very difficult to automatically analyze microarray images due to several problems such as spot position variation. We propose a novel block and spot indexing algorithm with the use of maximal epsilon-regularity. The time complexity of our algorithm is O(n²) where n is the number of cells.

The method combines the tasks of classification and feature selection by using the obtained clusters in ECOS to further extract specific features for each of the clusters. The method overcomes the problem of the signal to noise ratio method when data of the same class are spread in several clusters.

The availability of entire genomic sequences enables matching the short probe sequences of oligonucleotide arrays to their annotated gene representations. Here, we present a framework for estimating the sensitivity and specificity of gene-representing probe sets, and for integrating this information with the comprehensive genome and transcriptome repositories of the GeneCards databases suite.

An intrinsic weakness – spot inhomogeneity – can be turned into a strength by using pixel correlation methods to reliably extract red/green ratios, identify dye-selective quenching and cull spots by data quality. We compare our methods to established approaches.

Custom oligonucleotide array design is aimed at effectively interrogating a largest possible non-redundant set of transcripts under a physical size constraint. 200,000+ publicly available and proprietary mouse transcript sequences were clustered. The custom chip was subsequently extensively used to profile the expression in 70+ different tissues.

The user-friendly R-based software program NOM-R, implemented Yang et al’s normalization procedures is developed. This program is not only convenient for the biologists to use in microarray data normalization, but also can handle the repeated intensity values and dye-swap experiments, simultaneously.

To identify transcription factors that may play an important role in the differentiation of stem cells to neurons, high throughput gene expression experiment and computational analysis were performed. Our result suggest many transcription factors- novel transcription factors as well as those previously known to be involved in differentiation signaling.

We used a randomized Latin square design for blocking experimental variations and for estimating tamoxifen effect in human breast cancer cell lines assay. Orthogonal decomposition of a gene expression map based on the experimental design is expected to elucidate treatment effects as well as systematic error components.

We present several methods (qualitative and quantitative) for determining the quality of hybridizations to Affymetrix GeneChips. In particular, we make an attempt to quantify the effect of hybridization artifacts on estimates of gene expression.

We focused on histopathological phenotype prediction in colorectal carcinoma from microarray expression data, and investigated statistically their prediction performance of various combinations of classification techniques and gene selection methods. The results demonstrated detected marker genes and prediction accuracy strongly depends on the employed combination.

To automate the microarray image analysis, several processes were developed for detecting spots, filtering bad spots, and generating HTML reports. It can analyze a lot of microarray images without user’s attention. Therefore, it can be a connection in high throughput pipeline.

In cDNA microarray, the conversion of the amount of fluorescence to image intensity with the scanning process must be carefully handled to find the gene expression ratio. We developed a reverse scanning method for the microarray image and applied robust M regression to normalize the data from the compensated image.

OligoDesign is a webservice for the design of DNA/LNA mixmer oligonucleotides. The OligoDesign software features recognition and filtering of the target sequence by genome-wide BLAST analysis. It includes routines for prediction of melting temperature, self-annealing and secondary structure for LNA substituted oligonucleotides. The OligoDesign program is freely accesible at http://lnatools.com/

By applying the double clustering with the agglomerative information-bottleneck method to NCI60 cell lines, the correspondence between gene expression patterns and the ostensible origins of the tumours was verified. By computing 'entropy', mutual information, and its variation for several stages of clustering, an appropriate number of clusters could be estimated.

We describe a powerful method for comparison and visualisation of gene expression data from different microarray platforms. Co-inertia analysis (CIA) is a multivariate method that identifies co-relationships in multiple datasets. The genes from each dataset, which define these trends, can be identified. Further details: http://bioinfo.ucc.ie.

The false discovery rate (Benjamini & Hochberg, 1995) is widely used for multiple comparison problems, including gene expression array studies. For independent, but not necessarily identically distributed test statistics, we derive the joint probability distribution of the number of total and false rejections, and thereby provide methods for exact small sample power and sample size.

Improvements in quantifying spots can directly impact the identification of genes critical to the development and progression of cancer. We utilize a stellar photometric model, the Moffat function, to analyze cDNA membrane microarray images. In the current setting, we fit the Moffat function to cDNA spots on the array.

A multi-objective genetic algorithm based approach, combining univariate and multivariate techniques, was proposed to find optimal gene sets for sample classification on gene expression data automatically and unbiased. Eight genes with 93% LOOCV accuracy of KNN were selected to be the optimal predictive gene sets for Colon cancer dataset.

MGraph is a MATLAB toolbox, which applies graphical models to solve problems in microarray data analysis. MGraph with its graphical interface allows user to predict genetic regulatory networks by a graphical gaussian model, and to quantify the effects of different experimental treatment conditions on gene-expression profiles by graphical log-linear model.

SPEXS is an algorithm that extracts statistically overrepresented patterns in a set of sequences. SPEXS output generally contains too many significant patterns for a user to survey in detail, hence a post-processing step designed to group these patterns into key clusters is helpful. We present approaches to isolate these clusters.

Our laboratories have been keenly interested in infectomic analysis of transcription profiles in human BMEC infected with C. neoformans. We performed a time-course study of C. neoformans infection using oligonucleotide microarrays. We have found dynamic changes in transcription profiles of cytokines that are important for pathogenesis of Cryptococcus meningitis.

HyBrow is a prototype computer system comprising an event-based ontology for biological processes, an associated database and programs to perform hypothesis evaluation using a wide variety of available data. We demonstrate the feasibility of HyBrow, using the galactose metabolism gene network in Saccharomyces cerevisiae as our test system, for ranking alternative hypotheses in an automated manner.

Using site/motif Gibbs sampling, transcription factor binding sites can be found by analysing either the promoter sequences of co-expressed genes or the promoter sequences of orthologous genes. Tree-Gibbs sampling combines both data sources in one algorithm. This way additional information is available, resulting in improved, both in speed and accuracy, motif finding

Abbreviated Abstract The KBRIN bioinformatics core is attempting to create a Kentucky-wide network of bioinformatics expertise. This venture has led to the identification of knowledge- and compute-based resources. The core seeks to improve bioinformatics knowledge through research and the creation of bioinformatics courses, certificates, and degrees. The core web site is: http://www.kbrin.louisville.edu/about/bioinform_core.html

Genomics Institute of the Novartis Research Foundation (GNF) is developing data analysis and visualization tools on top of a web-based informatics system for its lead discovery biomedical research. Recent tools include: dose-response data fitting and visualization, structural similarity-based compound hierarchical clustering, ring component-based compound diversity analysis, LCMS data visualization, etc.

We present the integration of web services into Delta2D, an end-user 2D gel analysis application. With Delta2D's web services plugins, users can access data from external sources and display it alongside the image analysis results in a uniform way. We describe the plugin architecture and first experiences in its use.

Theorem proving is a method involving logical reasoning and has a variety of applications, including diagnosis and decision-making. Resolution refutation is a general technique of proving the validity of a theorem given a set of axioms and rules. To prove theorem proving, DNA molecular reaction is implemented by resolution refutation.

Version 2 of the G-language Genome Analysis Environment is developed to gain further speed and integrity through the object-oriented architecture. Powered by the front-end "inspire" based on Flash/HTML and the database system "bluebird", the new environment aims to achieve greater flexibility and integrity.

This paper describes integrated protein structure modeling by using a fact constellation model and represents this modeling to XML in order to store and query highly complex protein data based on a XML mediated warehouse system. It performs complex queries employed during analyzing process by using XML query processing.

Usually, personal computers use the MS Windows operating system, but the computing power is used for simple work. We try to use these available resources as member resources of a clustering system. XML-RPC provides the straight forward way for distributed BLAST and heterogeneous operating systems to be used as a member of the distributed system for BLAST.

MineLink is a novel information integration framework that can pull together life sciences data and analytic applications from disparate sources. It specifies a design methodology for automatically integrating individual components, be they data sources, processors, data miners or visualization components without the need for explicit programming. It addresses both syntactic and semantic aspects of information integration.

We developed TransMiner, a symbolic computational approach for inferring novel biochemical functions. We have developed a novel sub-graph isomorphism-based algorithm to search the detailed representations of known biotransformations to induce biocatalytic "rules". These symbolic biocatalytic rules represent the simplified functions of enzymes and can be used to infer novel biochemistry.

The poster shows a gene network based on a map of interactions between proteins and genes in Drosophila melanogaster. The map is based on cross correlations of genome wide time courses of gene expression and yeast to hybrid interactions. Results show a new rich network of connections between genes and proteins.

^myGrid aims to exploit Grid technology & provide high-level services and middleware that make it suitable for bioinformatics. ^myGrid uses resource discovery and workflow enactment services that allow scientists to perform in silico experiments over bioinformatics resources. Services are provided to support the scientific method, notably provenance management, change notification & personalization.

It is important to infer positive selection sites associated with a given gene family. Different approaches have been proposed to detect positive selection at single amino acid sites. The performance of these approaches was evaluated so that researchers should be able to apply an appropriate method to their research when certain circumstances are met.

To overcome the limitations of comparative mapping using orthologous genes, we have developed a new paradigm and systematic approach to define conserved synteny between human and mouse directly from genomic sequence. The automatic pipeline should be applicable to compare any species with complete or draft genome sequence and within an appropriate phylogenetic distance.

We analyzed locations of HERV-K LTRs in the human genome. Their distribution is not uniform among human chromosomes and among different regions of one chromosome. The majority of HERV-K LTRs are clustered. Positions of LTR clusters correlate with the Giemsa segmentation of human chromosomes. Most clusters are observed GC-rich regions.

Bang-Poong species is one of the most important species of herbal medicine. We have applied a molecular approach to developing SCAR markers. We show that this methodology can be applied to dried herbal medicine.

Maximum Likelihood methods are popular for analysing sequences to detect adaptive evolution. We have designed new methods for robustly inferring the evolutionary history of extant sequences and identifying signatures of adaptive evolution, performing analyses of multiple sequence alignments in ways closer to biological reality than existing methods. Available at http://bioinf.may.ie/likelihood.

We describe a method for modeling changes in codon substitution in populations where evolution can be measured, e.g. rapidly evolving viral populations. Our model extends Neilsen and Yang's codon model to allow parameters associated with selection and proportion of selected sites to vary over time.

Phylogenetic analysis of Toll gene evolution across insects, plants, and mammals using their transcriptional regulatory regions has been performed. This analysis is based on a unique approach employing the frequency of “evolutionarily” informative transcription factor binding sites. Interestingly, the resultant phylogeny produced results similar to that of protein-based phylogeny.

PyPop (Python for Population Genetics) is a suite of programs for the analysis of multi-locus population genetic data, outputs are stored in XML and can be transformed into other data formats. PyPop will be made freely available under the GNU GPL at: http://allele5.biol.berkeley.edu/pypop/

The history of China is a record of migration, population admixture and community endogamy. While comparing current Bayesian and "Classical" methodologies, these factors were prevalent in the genetic structure of eight Chinese ethnic populations. It’s concluded that Bayesian models comprising multiple-system data and historical factors are required for future studies.

We develop a hybrid approach to recognizing protein families among multi-genomic datasets based on Markov and single-linkage clustering of normalised pairwise BLASTP bit scores. We present results for all proteins from 114 microbial genomes, and illustrate its utility by recognizing orthologs and paralogs of rotary motor ATP synthetase F1 subunits.

DCM-GRAPPA is a method for phylogeny recontruction based on gene-order data; it scales gracefully to one thousand genomes, using a day or two of computation and producing highly accurate results (within 1%). GRAPPA and DCM-GRAPPA are available in source form at http://www.cs.unm.edu/~moret/GRAPPA/

The number of genes and species for which data are now available present an opportunity for statistically powerful examinations of molecular evolutionary processes. We will present a description of the functionality and performance of EVOLVE (cbis.anu.edu.au/software), a high performance computing package for phylogeny-based maximum likelihood modeling and hypothesis testing.

We have constructed a novel Markov model of dinucleotide substitution including parameters for methylation and strand using the EVOLVE toolkit. Our analysis supports a significant contribution of differential methylation in the germline elevating male mutation rates: clearly demonstrating the utility of EVOLVE (http://cbis.anu.edu.au/software/) in developing new sequence evolution models.

Alignments display a mosaic pattern due to recombination. Existing programs use a broad window to find major events. This leads to not recognising the finer detail where multiple events occur at different points in different lineages. BOSS uses pairwise comparison and a short sliding window to analyse this mosaic structure.

The relationship of nematodes to arthropods and vertebrates can be described by the Coelomata and Ecdysozoa hypotheses. Our aim was to test these hypotheses by finding the supertree that best described the relationship of orthologous, single gene family trees from ten eukaryotic taxa. Our results support the traditional Coelomata hypothesis.

Whole genome duplication has played a major role in the structuring of the Arabidopsis thaliana genome. We examined single nucleotide polymorphism (SNP) variation in duplicate genes retained after these duplication events. We compare SNP accumulation in duplicates and singletons with respect to their effect on protein evolution.

The classes AAC, AAG present in higher number in Arabidopsis, whereas AAT AGG and CCG are more abundant in rice. Repeats with A and T bases (except AAT) are more frequent in Arabidopsis, repeats with G and C bases are more in rice, as rice is high in GC composition.

DNA modification plays important roles in nucleic acid metabolism. Methyltransferases modifying the related DNA sequences GATC and GANTC are close homologs. We present a phylogeny of these enzymes as an example of how DNA sequence recognition has evolved, and predict how to evolve further specificities in vitro

To free phylogenies from alignments we present two approaches based on pattern discovery using TEIRESIAS: the first one computes distances from patterns, where distance is defined analogous to distances on alignments. The second approach transforms patterns into character data, meaning that we do not have to explicitly extract relevant properties.

Individual genes often fail to reproduce similar trees, sometimes with disparities so serious as to question the entire concept of organismal phylogeny for prokaryotes. However, trees produced from bulk genomic signal display topologies that are largely consistent with accepted taxonomy, suggesting that the underlying mode of prokaryotic evolution is clonal.

By analyzing the genome of Oryza sativa (rice), we show that a substantial fraction (15%) of all rice genes is found in duplicated segments. However, detailed analysis shows that rice is not an ancient polyploid as previously suggested, but an ancient aneuploid that has experienced partial genome duplication, approximately 70 million years ago.

Cathepsins are proteolytic enzymes of lysosomal origin. According their catalytic mechanism they can be classified as cysteine, serine or aspartic protases. In this work we investigated the phylogeny and evolution of human cathepsins.

EVA (Enteric Virus Analysis) is an analysis tool that brings together existing and novel computational techniques within a single integrated software environment. Here we describe the design of EVA and its use in examining the evolution of emerging group of foodborne viruses.

LumberJack is a phylogenetic tool intended to facilitate sampling treespace to find likely tree topologies quickly and to map phylogenetic signal onto regions of an alignment in a heuristic manner.

Microsatellite technology rapidly reveals high polymorphic fingerprints and thus determines the genetic markers. In combination with oligonucleotides of arbitrary sequence, 5' anchored oligonucleotides based on simple sequence repeats were used in PCR to produce Arisaema DNA fingerprints.

Classification of virus risk types is important to understand the mechanisms in infection. We propose a machine learning approach to classify HPV risk types. Our approach is based on the kernel method that provides efficient computation. In our experiments, the string kernel-based classifier predicted four unknown HPV types exactly.

We use the novel IL28A,B and IL29 protein family to illustrate an approach to the computational identification and characterization of putative transcriptional regulatory regions. Our technique consists of a combination of phylogenetic footprinting and detection of statistically significant clusters of binding sites and can be applied on a genomic scale.

An approach was developed for predicting the presence of a specific functional effect for active peptides. It involved multiple steps: a) collection of protein sequences from multiple sources, b) data cleaning and functional annotation, c) definition of basic structure-function unit groups, and d) prediction of protein function by an intelligent agent.

Nuclear Receptors (NR) control diverse programs of gene expression. These transcription factors bind in homo- and heterodimeric forms to complex target sequences. Due to variable spacing and orientation of half-sites, standard profile models are inadequate. We construct an HMM framework for the prediction and classification of NR binding sites.

The action of the combine therapy of chloroquine,puritine and ascobic acid has baffled biochemist for some times.but it is believe to potentiate its action based on the pertubation of the lipid bilayer surrounding the cells. but further studies to reveal the genetic conditions has not yielded much results.

The ability of each and every organism to adapt its particular environmental niche is of fundamental importance to its survival and proliferation. We have identified species-specific protein sequence and fold optimizations, which we exploited to generate predictive scoring functions. These scoring functions will be used in future species-specific protein identification and optimization experiments.

A genetic algorithm has been applied to predict building schemas of protein secondary structure. Although the average Q3 of this research is not the highest score among researches, some fundamental and useful building schemas of protein secondary structure have been found.

The ability of paullones and aloisines to inhibit glycogen synthase kinase-3NULL and cyclin-dependent kinases was well predicted by 3D-QSAR using CoMSIA in conjunction with a novel alignment method based on electrostatic potentials. The predictive value of this approach may lead to the development of better therapeutics for Alzheimer’s disease.

REMmatch program is designed for preliminary screening of any sequence database for searching potential hormone responsive elements (HRE) and is comparable with other known programs TESS, Match etc.). REMmatch is available at http://www.math.wisc.edu/~karp/REMmatch.exe

We suggest a new algorithm that allows prediction of additional regulatory functions for the proteins containing different structural motifs. This algorithm was successfully applied for a set of proteins containing ubiquitin motifs where we could also show their regulation by certain protein kinases.

We successfully predicted synthetic lethal gene pairs in Saccharomyces cerevisiae. Using probabilistic decision trees, we integrated multiple data types including correlated mRNA expression, physical interaction, protein function, and sequence homology. Our predictions may help identify redundant genes and pathways and may may better our understanding of genetic robustness.

We present PSORT-B (http://www.psort.org) - a subcellular localization predictor with a measured accuracy of 97% for Gram-negative bacteria. Issues including the handling of proteins resident at multiple localizations, the importance of the “unknown” result and the development of a Gram-positive version are discussed. Selected whole genome analysis is also presented.

The nucleus belonging to any type of the cells was determined to estimate of atypical lymphocytes in patients with chronic tonsillitis, to analyze the vital cycle those cells on the base of classification of transformed cells stained according to Romanovsky-Gimza. Image segmentation process was adapted to condition of low-budgetary science

MicroRNAs are a class of tiny RNA genes excised from precursor hairpin structures. We identified a highly specific conserved motif upstream of precursors that might be involved in miRNA transcription, and describe how much this motif, plus features of general upstream and downstream conservation, aid in miRNA gene finding.

Modelling large bio-molecules is still primarily limited to the simulation of short timeframes, which in many cases are not biologically significant. The goal of this project is to use the optimisation of Hamiltonian paths to enable calculation of the behaviour of molecular systems over large time frames.

A bioinformatics approach is presented that can bypass in some instances the traditional means of phosphopeptide mapping a radiologically labeled target protein to identify PKC phosphorylation sites. This directed bioinformatics approach utilizes comparative sequence analysis, molecular modeling, and machine learning techniques to assist in the discovery of PKC phosphorylation sites.

We define a new class of pseudoknots and present algorithms for thermodynamic folding of RNA secondary structures including such pseudoknots. Their time/space complexity is O(n^4) and O(n^2). We also compute the best structure guaranteed to contain a pseudoknot, and the most tightly knotted substructure. An extensive evaluation of Pseudobase is performed.

Our strategy is based on the biological fact which show that promoters are localized in the upstream regulatory regions of genes. The possibility for a particular sequence to be a bona fide promoter can be evaluated from its mismatch distribution amongst the various areas of the genome.

NetOGlyc 3.0 is a predictor of mucin type O-glycosylation sites, predicting from the protein sequence alone. NetOGlyc 3.0 shows much better generalization behaviour (the ability of the network to correctly predict for completely new examples) than its predecessor and is available at www.cbs.dtu.dk/services/NetOGlyc/

We find differences in the distribution and conservation of vertebrate splicing regulatory elements and relevant trans-factors given the availability of large-scale genomic data for Homo Sapiens, Mus musculus and Fugu rubripes.

We investigated on the novel use of Amplified Fragment Length Polymorphism screening in the diagnosis and classification of cancers using a set of 58 gastric tumor and 16 normal genomic DNA samples. The result shows that SVM can be used to differentiate cancer from non-cancer tissues with high accuracy.

In order to annotate the membrane organization of whole-proteome datasets, we have developed a consensus annotation protocol automated for high through-put analysis. This protocol predicts the presence of transmembrane domains, gpi-modifications and signal peptides. These features are combined to generate a prediction of membrane organization.

We mimic the major geometric features of the native structures of globular proteins: compactness, hydrophobic core, and smaller surface area. We hypothesize that the native structure of a globular protein should minimize all above three geometric features simultaneously and coherently among all conformations satisfying a relaxed steric condition.

We report an analysis of melting curves for genomic DNA. Our in-house software employs novel estimates for the weights of loop entropy factors. As test-cases, we studied D. Discoideum and synthetic sequences inserted in a linearized plasmid to compare with experiment. We find that the cooperativity parameter may be one order of magnitude larger than its consensus value.

Ubx and Exd are homeodomain transcriptional factors in Drosophila that cooperatively bind DNA to regulate cell differentiation. Using FADE, we discovered an interaction between these proteins that is nearly absent in the human homologs. Through chemical analysis we find that this interaction may increase binding in fly, but don't expect this in human.

Outer membrane proteins (OMPs) of bacteria are medically important as drug targets. We developed two OMP predictors based on frequent subsequences studied in data mining. Both significantly outperformed the state-of-the-art method and one also produced explicit patterns of OMPs that can be used for further biological analysis.

This article presents a hybrid method that can identify the individual's haplotype from the given genotypes. Our method combines statistical and computational approaches in order to increase the accuracy. The individuals' haplotypes are resolved by considering the MLE (maximum likelihood estimation) in the process of computing the frequencies of the common haplotypes.

We predict snoRNAs in human genome using several methods for sequence analysis. We also develop a Predicted Human Intron database produced from exons predicted by Gene Decoder which is a gene finding technology based on HMMs. We show the results of prediction of snoRNAs and the databases of human introns.

We developed the system, PIVS, which is very useful for predicting the function and interaction of unknown protein sequence as well as for visualizing its protein-protein interaction map. In addition, it offers integral genomic and motif/domain-related information concerning unknown input protein sequence.

We present a novel correlation based technique for unsupervised feature extraction in large datasets. The algorithm selects features based on their redundancy and unlike PCA, preserves the spatial information in the data, allowing one to easily interpret the extracted features. We demonstrate that classification accuracy on the reduced feature data is comparable to that on the original data.

We have used support vector machines and Naive Bayes methods to classify protein surface residues into interface residues and non-interface residues based on the sequence neighbors of target residues. The results showed that both methods are able to successfully discover and use sequence neighbor features predictive of functional properties to identify interface residues.

The RIKEN Representative Protein Set was subdivided into groups based on signal peptide and transmembrane domain combination. Functional analysis using InterPro and SCOP was performed on three sub-groups: type I and II transmembrane and secreted proteins. The sub-cellular localization of type II proteins were computationally predicted and tested in vivo.

We describe a novel statistical model of protein sequences and interaction networks that utilizes discriminative and informed priors, and apply it via a Gibbs sampling algorithm to the experimental SH3 domain-peptide interaction network of Tong et al (2001). Our results reveal that such interaction networks can, to a large degree, be modelled by this technique.

We have attempted neural-network-based predictions of inhibition coefficient (IC50) from the SMILES of ligands for kinases and proteases in the protein-ligand interaction database, ProLINT. Our method is useful for a large scale filtering of ligands in drug-design. We have also attempted to develop ligand fragment-signature for proteins in ProLINT.

EPP is a eukaryotic promoter prediction system. In EPP, after training set is divided into many clusters, each cluster is separately trained to make a decision model. This approach enhances the sensitivity and specificity of EPP. EPP is available at http://epp.cau.ac.kr

Single base extension (SBE) is the most widely used SNP genotyping method. When multiplex SBE (MSBE) is applied to highly polymorphic regions, hybridisation to variable DNA may occur. We have developed an MSBE primer design algorithm that increases the coverage of MSBE for these regions by up to 20%.

Current genotyping analysis packages provide basic genetic marker interpretation. However, many applications often require further manual specialised analysis, which is labour intensive and expensive. We developed Genefiler, software providing massive genotyping capability, giving diagnostic results from raw data, comprehensive project management, intuitive GUI and a flexible modular format.

We have analyzed relationship between structure and function (activity) of transcription factors, based on two approaches: a knowledge-based approach, utilizing structural data of protein-DNA complexes, and computer simulations. We have examined the roles of structural deformation of DNA and cooperativity in target recognition, and predicted targets in yeast genome.

Peptide mass fingerprint (PMF) has been a useful method for rapid and high-throughput protein identification. In our study, we compared software used frequently to identify the proteins of Homo sapiens and Halobacterium salinarum. These attempts could provide more effective algorithm for protein identification of each species using PMF.

We describe a general dynamic programming-like algorithm (MaxSubSeq) specifically designed to optimise the number and length of segments with constrained length in a protein sequence. Our algorithm is independent of the underling predictive method and is available through the web interface at http://gpcr.biocomp.unibo.it

We have developed a Web based system for searching transcriptional regulatory elements in Human Genome Sequences using Transcription Factor Database (TFDB: which we have been maintaining) and Genome comparison. Combined with Microarray data, we will give some common regulatory elements between similar expression genes.

This poster presents PathMiner, a computational framework for exploring metabolic pathways. To make inferences about pathways we abstract metabolism as a state-space in which compounds are points and biotransformations are state-transitions. In this poster, we discuss applications of PathMiner to two quite different biological problems.

Single nucleotide polymorphism (SNP) distribution has been demonstrated as nonrandom in the genomes of animals and plants. While genetic variation is traditionally expected in noncoding regions, additional genetic feature have been correlated with SNPs. We investigate whether this information might be used to predict regions of polymorphism.

Using maximum likelihood estimation, we analyze the correspondence between popular scoring matrix classifiers for binding site prediction and specific probability distributions of binding site sequences. For unknown distributions, the binding matrix is a good choice since it achieves maximal specificity under the constraint that all known sequences are classified correctly.

CpG-islands are considered evolutionary remnants linked to the regulation of genes. Compared to animal systems, plants have a higher number of genes encoding DNA-methyltransferases that show a broader specificity. In this work we explored the compositional landscape surrounding promoters that would enable the identification of distinct CpG or CpNpG-islands.

The predictive performance and comprehensibility of 54 chemical-carcinogenicity models, generated by three Predictive-Toxicology Challenges, was evaluated using ROC convex-hull analysis. Models from problem representations that used a mix of attributes, reflecting interactions between chemical and biological features, clearly outperformed those derived using only attributes of chemical structure or biological features.

Comparative computational analysis can lead to improved identification of genes, while relying on a given pair of genomes, such as human and mouse. We propose a formal way to select an optimal pair of genomes by linking Markov models of molecular evolution to comparative HMMs and studying their prediction accuracy.

There is a weak but significant similarity between the prion protein (PrP) and some transcription factors and Zn-finger proteins. A molecular model of the Cu++-binding monomeric (normal, cytoplasmic) PrPC and the Cu++-stabilized polymeric (scrapie - pathogenic) PrPSc is presented and is called the CUPRION model.

In some applications of sequence comparison theories the actual items to be compared are not successions of discrete elements, but "continuous" functions of a continuous argument. The present paper is aimed to construct a "continuous" distance function with the help of the given "distance" matrix D.

We derived and tested novel species-specific substitution matrices (SSSMs) for sequence alignment. Our results show increased alignment accuracy in the 20-30% sequence identity range, but decreased alignment length as compared to popular sequence alignment programs such as PSI-BLAST (using BLOSUM) using CASA’s SCOP based alignment test sets.

We have analyzed genetic and amino acid residue variation of Plasmodium falciparum cell surface antigens (merozoite surface proteins 1 and 2; circumsporozoite protein; stevor and rifin) in the context of their known or predicted secondary structures. Locations of structural motifs suggest the presence of functional domains or antigenic epitopes.

Finishing a genome costs as much as initial assembly. Since initial assembly gets about 95% of it, gaining just a few extra percent initially can save tens of millions of dollars in finishing costs. We present computational techniques for improving "overlaps" result in up to 5% additional sequence during initial assembly.

The Bielefeld University Bioinformatics Server (BiBiServ), http://bibiserv.techfak.uni-bielefeld.de, supports Internet-based collaborative research and education in bioinformatics. Currently, 15 software tools and various educational media are available. These include tools from different areas such as Genome Comparison, Alignments, Primer Design, RNA Structures, and Evolutionary Relationships. In 2002 approximate 14.000 users per month used BiBiServ services with rising tendency.

G1/S checkpoint plays a pivotal role during early stages of the cell cycle. Progression through G1/S boundary is controlled by a series of regulators. To explore the roles of these regulators and identify targets of CDKs, bioinformatics analysis are being done on DNA and protein sequences from different eukaryotic organisms.

We developed a rapid approximate pattern matching algorithm and a linear time algorithm for multiple alignment construction, with no previous pairwise matching of sequences required. These are implemented in a program for shotgun sequence error correction able to correct 99% of sequencing errors. This is significantly better than previous methods.

A parallel HMM-pfam is implemented on EARTH - an event-driven fine-grain multi-threaded program execution model. It demonstrated significant performance improvement over another PVM based version. On a cluster of 128 dual-CPU nodes, the execution time of a representative testbench is reduced from 15.9 hours to 4.3 minutes.

We used coexistent genome-wide unique sequences in human and mouse genomes to recognize their homologous regions. The resulting syntenic map revealed more than 400 conserved segments and covering more than 90% of both genomes. This alignment-free method is capable of comparing two mammalian genomes within hours on one personal computer.

Long-range correlations in the amino-acid sequences of natural proteins are extracted from GanBank sequences. Some bigrams with more than one letter gaps turned out to be clearly over-represented. This data may improve the assessment of alignment quality, phylogeny analyses, and perhaps database searches.

The Smith-Waterman algorithm is a dynamic-programming algorithm that finds the optimal alignment between two biological sequences. The algorithm was implemented on a field programmable gate array (FPGA) in order to investigate the advantages and disadvantages of using reconfigurable computers and associated tools for computationally intensive sequence analysis applications.

T-Coffee is a novel method for multiple sequence alignment that allows you to combine heterogeneous sources of data to produce very accurate alignments. In this poster we look at the effects of mixing sequence and structural information using two structural alignment programs, SAP and FUGUE, in combination with T-Coffee.

Correlation between activity of antisense oligonucleotides and local structural features of target RNAs is studied. Statistical analysis showed that high activity of antisense oligonucleotides is more likely to occur in regions of target RNA having hairpin or multi-branched loops with flanking stems.

To better align EST and genomic sequence, we developed a tool named Alexa, which incorporates a splice site model into the recursive dynamic programming equation. For reduced memory consumption and increased speed, Alexa can be guided by an input file produced by WU-BLASTN. Alexa is available at http://sapiens.wustl.edu/alexa.

ASAD is A Sequence Attribute Display tool. It is written as a set of Excel macros. It builds on the familiar Excel interface to allow for flexible and efficient display of attributes (hydrophobicity etc.) using similar colours and styles for one sequence or multiple aligned sequences. ASAD is available at ftp://ftp.wehi.edu.au/pub/biology/ASAD.

PSI-BLAST is a sensitive and fast database search algorithm, however false positive results can taint the final results. We are currently investigating ways of detecting when the matrix generated by PSI-BLAST no longer describes the original query sequence or family.

We aim to identify putative regulatory elements in the orthologous human promoters of co-regulated bovine genes. Promoters were extracted and Motif prediction was performed by MEME. The MatInspector Professional software and BioPerl programs were used to assess the results. Novel putative elements were identified in the human promoters.

LAGAN2 is a probabilistic method for limited area global nucleotide alignment of distantly related species. The algorithm incorporates multiple conservation models for protein-coding, non-coding, and unconstrained alignments, approximate logarithmic gap penalties, and Hidden Markov Model based training of alignment parameters through expectation-maximization. Additional information is available at http://lagan.stanford.edu.

Frequency enumeration of the DNA subsequence is one of the important techniques. The algorithm is simple, but it takes enormous memory space. We propose an enumerate method that uses linear codes to reduce the memory space. The method with 2-ary (31,26) Hamming code requires 1/60 of the usual memory space.

SeqFreq is a repetitive motif discovery tool for finding repeats in both intragenomic and intergenomic sequences. SeqFreq uses a numerical suffix tree method to enumerate repetitive n-mers. Currently, intragenomic n-mer repeats of multiple bacterial genomes and their statistical distributions have been analyzed.

A novel method called DSC (Difference String Comparison) is proposed for speeding up the primer finding procedure. DSC method presents a partially ordered graph of DNA sequences and also reserves all information of sequences.

ClustalW analysis of 27 corona virus genome sequences reveals that 9 SARS virus sequences are identical. Multiple sequence alignment was also performed to the different groups of corona virus genome sequence and coding product to find conservative and divergent regions. All the analysis results are available at ftp://ftp.cbi.pku.edu.cn/pub/sars/analysis/.

An alternative method to TblastX has been developed, known as blastNP. Nucleic acids in database and query sequences were translated into overlapping protein-like sequences (overlappingly translated sequences, OTSs) before searching with blastP. Thus, each nucleic acid sequence is represented by a single “protein like” sequence (instead of three reading frames).

An analysis of the quality of the profile--profile alignments generated by a BASIC-like algorithm highlights that Shannon entropy can be used to filter out most of the bad high-scoring alignments, enhancing its reliability in the detection of remote homology. When entropy-filtering is used, the best-scoring alignments are comparable to that obtained by the CE structural alignment algorithm.

An approach based on computational geometry is used to elucidate structural changes in a HIV-1 protease monomer caused by dimerization and inhibitor binding. A comprehensive mutational analysis of HIV-1 protease is also performed using this method and reveals a strong structure-function correlation.

We present here a new tool for the discovery of unstructured, or disordered regions within proteins. GlobPlot http://globplot.embl.de is a web service that allows the user to plot the tendency within the query protein for order/globularity and disorder.

Using a novel physics-based approach, complete protein unfolding pathways for five distinct folds were computed and compared with published molecular dynamics and in-vitro experiments. Agreement was good and computation reduced by three orders of magnitude, suggesting that unfolding pathways of small globular proteins are largely constrained by geometry and sterics.

We present a modified version of the TRADES algorithm for homology modeling of protein sequences given weak similarity to experimentally determined protein structure that generates realistic, all-atom models of non-idealized geometry as it incorporates backbone dependent rotamers, reasonable bond lengths, bond angles, torsion angles, minimized electrostatics and van der Waals forces.

We have proposed a novel parameter, long-range order (LRO) for a protein from the knowledge of long-range contacts in protein structure. LRO correlates very well with experimental protein folding rates. The short and medium-range non-bonded energy, long-range contacts, and helical/strand tendency are the major determinants for transition state structures of two-state proteins.

A homology model was constructed for the DHPS-PPPK bifunctional enzyme of Plasmodium falciparum. This enzyme plays a vital part in folate synthesis and is targeted by current, failing therapies. The LUDI/ACD and the NCI database was scanned against the models to identify new potential inhibitors.

Three dimensional circle fitting using atomic coordinate was performed to all known structures of leucine-rich repeat (LRR)-containing proteins. The analysis results indicate that there is a regular relationship between the radius of the LRR arc and the rotation angle about the central axis of the arc per repeating unit.

A hybrid algorithm combining genetic algorithm and tabu search is presented. The hybrid algorithm can be successfully applied to protein folding based on a hydrophobic-hydrophilic model. A protein structure database is also created. The results indicate that in all cases our method works better than genetic algorithm or tabu search alone.

In order to explore the relationship of protein function with secondary structural attributes, we compared simple statistical attributes of secondary structural elements in three major enzyme classes. We tested the usefulness of these inferences in predicting the enzyme functional class using neural networks. The strategy presented may help facilitate broad functional annotation of unknown proteins.

We developed a prediction method for protein interaction sites using sequence profiles and accessible surface area of neighboring residues, surface patches, other physicochemical characteristics, and support vector machines. The relationship between the prediction accuracy and the characteristic of protein-protein interaction sites is discussed.

Incorporating sequence and biochemical features in TOPS (Topological Models of Protein Structures) is significant for pattern matching and pattern discovery in protein structures. Interesting results would be valuable for efforts to predict protein structure and function from sequences. These problems remain key challenges of direct relevance to projects in structural and functional genomics. TOPS is available at http://www.tops.leeds.ac.uk

We present a bioinformatic toolbox for postprocessing of proteomic data obtained by peptide fingerprint analysis. This toolbox (i) increases the reliability of protein identifications, (ii) detects additional annotation information and (iii) corrects or validates start codon assignments by gene finders. The toolbox was developed for and applied to the proteome of Halobacterium salinarum strain R1 (http://www.halolex.mpg.d)

In the set of proteins of H. salinarum identified by MALDI-TOF peptide fingerprints, membrane proteins are severely underrepresented. Besides the technical problems encountered in their analysis by 2D gel electrophoresis, an ‘inherent data analysis problem’ is found by statistical analysis. The degree to which this effect aggravates problems in detection ratio of membrane proteins varies from organism to organism.

We present a fold recognition method that uses predicted secondary structure and solvent accessibility in a way that significantly improves both specificity and sensitivity of fold assignments compared to PSI-BLAST. The method can readily be used for high quality/throughput database annotations.

There exist several classes of protein structure that are determined by predominance of an element of secondary structure. Their recognition, starting from aminoacid composition, gives some insight into the nature of this classification. Class recognition methods are built by means of neural networks. The accuracy of recognition has reached 75%.

A computational approach is presented to explore relationships between functional specifity of proteins and hydropathy distribution in proteins. The approach uses features based on hydrophobicity of amino acids to model the hydropathy distribution in proteins. A feature space is employed to identify functional protein families and the features that are specific for the families.

We present a new approach to protein structure comparison using the existing, on graph representations based, TOPS database. Instead of comparisons based on single patterns, we use profiles of patterns, which allows in an indirect way to capture "negative" information. This leads to a significant increase in prediction accuracy.

A global view of the “protein structure universe” was constructed. Such a representation reveals a high-level organization of the fold space that is intuitively interpretable, offering an interesting perspective on both the demography and the evolution of protein structures. The protein fold space is available at http://pro.lbl.gov/~jingtong/foldspace.

IDPharmoTM integrates programs that can discover the potential lead compounds using protein structures and 3D chemicals database in silico. This tool was evaluated for discovery of potential lead compound of various disease targets such as HIV-RT, PDE5, HCV pol and DPP IV. We could get new chemical entities whose biological activities are from IC50 50uM to 1uM

We performed molecular dynamics simulations on monomeric and dimeric forms of human prion protein at various conditions (at 300K, 500K, acidic pH, or with a mutation D178N) for 10 ns to investigate the differences in the dynamics of each form. Most differences resulted from additional inter-subunit interactions of dimeric HuPrP.

Expeditious mapping of exon borders and intron phase data onto structurally similar proteins presents a novel approach to studying protein structure evolution. We present SEDB, which allows researchers to examine and quickly identify where exon/intron boundaries are located and how these borders have possibly shaped current protein structure. http://glinka.bio.neu.edu/~cleslin/SEDB/SEDB.html

The mechanism of protein-gated electron transfer between two quinones of photosystem II was investigated. Sequence and structural conservation of a high-packing motif including an intersubunit H-bond, combined with in-silico and in-vivo combinatorial mutagenesis and biophysical characterization of the H-bond donor suggests that reversible dissociation of the bond regulates the gating.

Identifying small molecules that modulate protein-protein interactions continues to be a major challenge for drug discovery. From a database of homologous protein-protein interactions, datasets were extracted representing the interaction region of pairs of proteins of these complexes. The aim of this research is to cluster features of each of these datasets.

S4E, a protein structure comparison system using constraint technology, searches common substructures of secondary structure elements between two proteins given in PDB format. For fast comparison, we developed an efficient algorithm for constructing the compatibility graphs and applying constraint programming to finding maximal common subgraphs.

The glucocorticoid receptor selectively recognises cortisol with high affinity. We build a homology model of this receptor on the basis of the progesterone receptor x-ray structure and simulated binding of different steroids (estradiol, progesterone, testosterone, aldosterone and cortisol). Using molecular dynamics simulations and free energy calculations we were able reveal the molecular basis of ligand recognition.

We propose an approach to predict the protein disulfide connectivity directly from the sequence. The proposed approach trained a SVR model and predicted the disulfide connectivity by Gabow’s algorithm. The experiments showed the proposed method has an accuracy of 62%, which is promising to locate the disulfide bridges in proteins.

XNAfold is a JAVA-C hybrid program that takes an RNA sequence as input and predicts RNA secondary structure. The minimum free energy structure predicted by XNAfold matched the experimentally determined structure for 77 out of 133 different RNA molecules. XNAfold is freely available at http://www.student.unimelb.edu.au/~yanmz/index.html

Beta sheets are one of the two common repeating elements found in protein structures. Despite their importance in structure they are not well or fully understood. Here we investigate the properties of beta sheets in order to better predict protein structure/folding and understand amyloid (and general aggregate) formation.

Of the 290 samples collected, only Campylobacter jejuni was isolated. Highest isolation rate was from dairy cows (54%), followed by urban sparrows (40%), farm sparrows (38%), rodents (11%) and flies (9%). Molecular charecterisation with PFGE of campylobacters provided 22 restriction patterns, of which 7 were common to more than one source.

We present a structural model of the SARS main cysteine proteinase (MPro), based on the X-ray structure of the transmissible gastroenteritis (corona)virus. From the SARS genome, peptides representing real MPro substrates have been docked into the active site. MPRo, with its docked ligands provides important clues for designing proteinase inhibitors.