Proceedings. International Conference on Intelligent Systems for Molecular Biology最新文献

The breakpoint phylogeny is an optimization problem proposed by Blanchette et al. for reconstructing evolutionary trees from gene order data. These same authors also developed and implemented BPAnalysis [3], a heuristic method (based upon solving many instances of the travelling salesman problem) for estimating the breakpoint phylogeny. We present a new heuristic for this purpose; although not polynomial-time, our heuristic is much faster in practice than BPAnalysis. We present and discuss the results of experimentation on synthetic datasets and on the flowering plant family Campanulaceae with three methods: our new method, BPAnalysis, and the neighbor-joining method [25] using several distance estimation techniques. Our preliminary results indicate that, on datasets with slow evolutionary rates and large numbers of genes in comparison with the number of taxa (genomes), all methods recover quite accurate reconstructions of the true evolutionary history (although BPAnalysis is too slow to be practical), but that on datasets where the rate of evolution is high relative to the number of genes, the accuracy of all three methods is poor.

The immense volume of data resulting from DNA microarray experiments, accompanied by an increase in the number of publications discussing gene-related discoveries, presents a major data analysis challenge. Current methods for genome-wide analysis of expression data typically rely on cluster analysis of gene expression patterns. Clustering indeed reveals potentially meaningful relationships among genes, but can not explain the underlying biological mechanisms. In an attempt to address this problem, we have developed a new approach for utilizing the literature in order to establish functional relationships among genes on a genome-wide scale. Our method is based on revealing coherent themes within the literature, using a similarity-based search in document space. Content-based relationships among abstracts are then translated into functional connections among genes. We describe preliminary experiments applying our algorithm to a database of documents discussing yeast genes. A comparison of the produced results with well-established yeast gene functions demonstrates the effectiveness of our approach.

We have developed a set of methods and tools for automatic discovery of putative regulatory signals in genome sequences. The analysis pipeline consists of gene expression data clustering, sequence pattern discovery from upstream sequences of genes, a control experiment for pattern significance threshold limit detection, selection of interesting patterns, grouping of these patterns, representing the pattern groups in a concise form and evaluating the discovered putative signals against existing databases of regulatory signals. The pattern discovery is computationally the most expensive and crucial step. Our tool performs a rapid exhaustive search for a priori unknown statistically significant sequence patterns of unrestricted length. The statistical significance is determined for a set of sequences in each cluster with respect to a set of background sequences allowing the detection of subtle regulatory signals specific for each cluster. The potentially large number of significant patterns is reduced to a small number of groups by clustering them by mutual similarity. Automatically derived consensus patterns of these groups represent the results in a comprehensive way for a human investigator. We have performed a systematic analysis for the yeast Saccharomyces cerevisiae. We created a large number of independent clusterings of expression data simultaneously assessing the "goodness" of each cluster. For each of the over 52,000 clusters acquired in this way we discovered significant patterns in the upstream sequences of respective genes. We selected nearly 1,500 significant patterns by formal criteria and matched them against the experimentally mapped transcription factor binding sites in the SCPD database. We clustered the 1,500 patterns to 62 groups for which we derived automatically alignments and consensus patterns. Of these 62 groups 48 had patterns that have matching sites in SCPD database.

Determination of the functions of all expressed proteins represents one of the major upcoming challenges in computational molecular biology. Since subcellular location plays a crucial role in protein function, the availability of systems that can predict location from sequence or high-throughput systems that determine location experimentally will be essential to the full characterization of expressed proteins. The development of prediction systems is currently hindered by an absence of training data that adequately captures the complexity of protein localization patterns. What is needed is a systematics for the subcellular locations of proteins. This paper describes an approach to the quantitative description of protein localization patterns using numerical features and the use of these features to develop classifiers that can recognize all major subcellular structures in fluorescence microscope images. Such classifiers provide a valuable tool for experiments aimed at determining the subcellular distributions of all expressed proteins. The features also have application in automated interpretation of imaging experiments, such as the selection of representative images or the rigorous statistical comparison of protein distributions under different experimental conditions. A key conclusion is that, at least in certain cases, these automated approaches are better able to distinguish similar protein localization patterns than human observers.

Sequence comparison of proteins that adopt the same fold has revealed a large degree of sequence variation. There are many pairs of structurally similar proteins with only a very low percentage of identical residues at structurally aligned positions. It is not clear whether these few identical residues have been conserved just by coincidence, or due to their structural and/or functional role The current study focuses on characterization of STructurally Aligned Identical ResidueS (STAIRS) in a data set of protein pairs that are structurally similar but sequentially dissimilar. The conservation pattern of the residues at structurally aligned positions has been characterized within the protein families of the two pair members, and mutually highly and weakly conserved positions of STAIRS could be identified About 40% of the STAIRS are only moderately conserved, suggesting that their maintenance may have been coincidental. The mutually highly conserved STAIRS show distinct features that are associated with protein structure and function: a relatively high fraction of these STAIRS are buried within their protein structures. Glycine, cysteine, histidine, and tryptophan are significantly over-represented among the mutually conserved STAIRS. A detailed survey of these STAIRS reveals residue-specific roles in the determination of the protein's structure and function.

The identification of sequence motifs is a fundamental method for suggesting good candidates for biologically functional regions such as promoters, splice sites, binding sites, etc. We investigate the following approach to identifying motifs: given a collection of orthologous sequences from multiple species related by a known phylogenetic tree, search for motifs that are well conserved (according to a parsimony measure) in the species. We present an exact algorithm for solving this problem. We then discuss experimental results on finding promoters of the rbcS gene for a family of 10 plants, on finding promoters of the adh gene for 12 Drosophila species, and on finding promoters of several chloroplast encoded genes.

In this paper we address the problem of reliably fitting parametric and semi-parametric models to spots in high density spot array images obtained in gene expression experiments. The goal is to measure the amount of label bound to an array element. A lot of spots can be modelled accurately by a Gaussian shape. In order to deal with highly overlapping spots we use robust M-estimators. When the parametric method fails (which can be detected automatically) we use a novel, robust semi-parametric method which can handle spots of different shapes accurately. The introduced techniques are evaluated experimentally.

Ontologies are specifications of the concepts in a given field, and of the relationships among those concepts. The development of ontologies for molecular-biology information and the sharing of those ontologies within the bioinformatics community are central problems in bioinformatics. If the bioinformatics community is to share ontologies effectively, ontologies must be exchanged in a form that uses standardized syntax and semantics. This paper reports on an effort among the authors to evaluate alternative ontology-exchange languages, and to recommend one or more languages for use within the larger bioinformatics community. The study selected a set of candidate languages, and defined a set of capabilities that the ideal ontology-exchange language should satisfy. The study scored the languages according to the degree to which they satisfied each capability. In addition, the authors performed several ontology-exchange experiments with the two languages that received the highest scores: OML and Ontolingua. The result of those experiments, and the main conclusion of this study, was that the frame-based semantic model of Ontolingua is preferable to the conceptual graph model of OML, but that the XML-based syntax of OML is preferable to the Lisp-based syntax of Ontolingua.

An efficient node-deletion algorithm is introduced to find submatrices in expression data that have low mean squared residue scores and it is shown to perform well in finding co-regulation patterns in yeast and human. This introduces "biclustering", or simultaneous clustering of both genes and conditions, to knowledge discovery from expression data. This approach overcomes some problems associated with traditional clustering methods, by allowing automatic discovery of similarity based on a subset of attributes, simultaneous clustering of genes and conditions, and overlapped grouping that provides a better representation for genes with multiple functions or regulated by many factors.

Mass spectrometry (MS) promises to be an invaluable tool for functional genomics, by supporting low-cost, high-throughput experiments. However, large-scale MS faces the potential problem of mass degeneracy--indistinguishable masses for multiple biopolymer fragments (e.g. from a limited proteolytic digest). This paper studies the tasks of planning and interpreting MS experiments that use selective isotopic labeling, thereby substantially reducing potential mass degeneracy. Our algorithms support an experimental-computational protocol called Structure-Activity Relation by Mass Spectrometry (SAR by MS), for elucidating the function of protein-DNA and protein-protein complexes. SAR by MS enzymatically cleaves a crosslinked complex and analyzes the resulting mass spectrum for mass peaks of hypothesized fragments. Depending on binding mode, some cleavage sites will be shielded; the absence of anticipated peaks implicates corresponding fragments as either part of the interaction region or inaccessible due to conformational change upon binding. Thus different mass spectra provide evidence for different structure-activity relations. We address combinatorial and algorithmic questions in the areas of data analysis (constraining binding mode based on mass signature) and experiment planning (determining an isotopic labeling strategy to reduce mass degeneracy and aid data analysis). We explore the computational complexity of these problems, obtaining upper and lower bounds. We report experimental results from implementations of our algorithms.