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

In this paper, the regulatory interactions between genes are modeled by a linear genetic network that is estimated from gene expression data. The inference of such a genetic network is hampered by the dimensionality problem. This problem is inherent in all gene expression data since the number of genes by far exceeds the number of measured time points. Consequently, there are infinitely many solutions that fit the data set perfectly. In this paper, this problem is tackled by combining genes with similar expression profiles in a single prototypical 'gene'. Instead of modeling the genes individually, the relations between prototypical genes are modeled. In this way, genes that cannot be distinguished based on their expression profiles are grouped together and their common control action is modeled instead. This process reduces the number of signals and imposes a structure on the model that is supported by the fact that biological genetic networks are thought to be redundant and sparsely connected. In essence, the ambiguity in model solutions is represented explicitly by providing a generalized model that expresses the basic regulatory interactions between groups of similarly expressed genes. The modeling approach is illustrated on artificial as well as real data.

In search for global principles that may explain the organization of the space of all possible proteins, we study all known protein sequences and structures. In this paper we present a global map of the protein space based on our analysis. Our protein space contains all protein sequences in a non-redundant (NR) database, which includes all major sequence databases. Using the PSI-BLAST procedure we defined 4,670 clusters of related sequences in this space. Of these clusters, 1,421 are centered on a sequence of known structure. All 4,670 clusters were then compared using either a structure metric (when 3D structures are known) or a novel sequence profile metric. These scores were used to define a unified and consistent metric between all clusters. Two schemes were employed to organize these clusters in a meta-organization. The first uses a graph theory method and cluster the clusters in an hierarchical organization. This organization extends our ability to predict the structure and function of many proteins beyond what is possible with existing tools for sequence analysis. The second uses a variation on a multidimensional scaling technique to embed the clusters in a low dimensional real space. This last approach resulted in a projection of the protein space onto a 2D plane that provides us with a bird's eye view of the protein space. Based on this map we suggest a list of possible target sequences with unknown structure that are likely to adopt new, unknown folds.

We present two algorithms which align flexible protein structures. Both apply efficient structural pattern detection and graph theoretic techniques. The FlexProt algorithm simultaneously detects the hinge regions and aligns the rigid subparts of the molecules. It does it by efficiently detecting maximal congruent rigid fragments in both molecules and calculating their optimal arrangement which does not violate the protein sequence order. The FlexMol algorithm is sequence order independent, yet requires as input the hypothesized hinge positions. Due its sequence order independence it can also be applied to protein-protein interface matching and drug molecule alignment. It aligns the rigid parts of the molecule using the Geometric Hashing method and calculates optimal connectivity among these parts by graph-theoretic techniques. Both algorithms are highly efficient even compared with rigid structure alignment algorithms. Typical running times on a standard desktop PC (400 MHz) are about 7 seconds for FlexProt and about 1 minute for FlexMol.

We have developed a method for the integrative analysis of protein interaction data. It comprises clustering, visualization and data integration components. The method is generally applicable for all sequenced organisms. Here, we describe in detail the combination of protein interaction data in the yeast Saccharomyces cerevisiae with the functional classification of all yeast proteins. We evaluate the utility of the method by comparison with experimental data and deduce hypotheses about the functional role of so far uncharacterized proteins. Further applications of the integrative analysis method are discussed. The method presented here is powerful and flexible. We show that it is capable of mining large-scale data sets.

External signals are transmitted to the cells through receptors activating signal transduction pathways. These pathways form a complicated interconnected network, which is able to answer to different stimuli. Here we analyze an important pathway for oncogenesis namely RAS/MAPK signal transduction pathway. We show that the interaction of the elements of this pathway induces topological structure in the element set and that the knowledge of the topology simplifies the analysis of the set. With a computer algorithm, we isolate from a large and complex group, smaller, independent, more manageable subsets, and build their hierarchy. Subsets introduction makes easier the search for attractors in discrete dynamical system, it permits the prediction of final states for elements involved in signal transduction pathways.

Understanding the mechanisms that determine the regulation of gene expression is an important and challenging problem. A fundamental subproblem is to identify DNA-binding sites for unknown regulatory factors, given a collection of genes believed to be coregulated, and given the noncoding DNA sequences near those genes. We present an enumerative statistical method for identifying good candidates for such transcription factor binding sites. Unlike local search techniques such as Expectation Maximization and Gibbs samplers that may not reach a global optimum, the method proposed here is guaranteed to produce the motifs with greatest z-scores. We discuss the results of experiments in which this algorithm was used to locate candidate binding sites in several well studied pathways of S. cerevisiae, as well as gene clusters from some of the hybridization microarray experiments.

We present a new approach for the evaluation of gene expression data. The basic idea is to generate biologically possible pathways and to score them with respect to gene expression measurements. We suggest sample scoring functions for different problem specifications. We assess the significance of the scores for the investigated pathways by comparison to a number of scores for random pathways. We show that simple scoring functions can assign statistically significant scores to biologically relevant pathways. This suggests that the combination of appropriate scoring functions with the systematic generation of pathways can be used in order to select the most interesting pathways based on gene expression measurements.

This paper presents a framework called Parallel Experiment Planning (PEP) that is based on an abstraction of how experiments are performed in the domain of macromolecular crystallization. The goal in this domain is to obtain a good quality crystal of a protein or other macromolecule that can be X-ray diffracted to determine three-dimensional structure. This domain presents problems encountered in real-world situations, such as a parallel and dynamic environment, insufficient resources and expensive tasks. The PEP framework comprises of two types of components: (1) an information management system for keeping track of sets of experiments, resources and costs; and (2) knowledge-based methods for providing intelligent assistance to decision-making. The significance of the developed PEP framework is three-fold--(a) the framework can be used for PEP even without one of its major intelligent aids that simulates experiments, simply by collecting real experimental data; (b) the framework with a simulator can provide intelligent assistance for experiment design by utilizing existing domain theories; and (c) the framework can help provide strategic assessment of different types of parallel experimentation plans that involve different tradeoffs.

Recent high-density microarray technologies allow, in principle, the determination of all k-mers that appear along a DNA sequence, for k = 8 - 10 in a single experiment on a standard chip. The k-mer contents, also called the spectrum of the sequence, is not sufficient to uniquely reconstruct a sequence longer than a few hundred bases. We have devised a polynomial algorithm that reconstructs the sequence, given the spectrum and a homologous sequence. This situation occurs, for example, in the identification of single nucleotide polymorphisms (SNPs), and whenever a homologue of the target sequence is known. The algorithm is robust, can handle errors in the spectrum and assumes no knowledge of the k-mer multiplicities. Our simulations show that with realistic levels of SNPs, the algorithm correctly reconstructs a target sequence of length up to 2,000 nucleotides when a polymorphic sequence is known. The technique is generalized to handle profiles and HMMs as input instead of a single homologous sequence.

The first available genome of a multicellular organism, C. elegans, was used as a test case for protein fold assignment using PSI-BLAST, followed by rational structure modeling and interpretation of experimental mutagenesis data in the context of collaboration with biologists. Similar results are demonstrated for human disease proteins with known polymorphisms.