Proceedings. IEEE Computational Systems Bioinformatics Conference最新文献

We have developed an ab initio algorithm for determining a protein backbone structure using global orientational restraints on internuclear vectors derived from residual dipolar couplings (RDCs) measured in one or two different aligning media by solution nuclear magnetic resonance (NMR) spectroscopy [14, 15]. Specifically, the conformation and global orientations of individual secondary structure elements are computed, independently, by an exact solution, systematic search-based minimization algorithm using only 2 RDCs per residue. The systematic search is built upon a quartic equation for computing, exactly and in constant time, the directions of an internuclear vector from RDCs, and linear or quadratic equations for computing the sines and cosines of backbone dihedral (phi, psi) angles from two vectors in consecutive peptide planes. In contrast to heuristic search such as simulated annealing (SA) or Monte-Carlo (MC) used by other NMR structure determination algorithms, our minimization algorithm can be analyzed rigorously in terms of expected algorithmic complexity and the coordinate precision of the protein structure as a function of error in the input data. The algorithm has been successfully applied to compute the backbone structures of three proteins using real NMR data.

A novel and rigorous Multi-perturbation Shapley Value Analysis (MSA) method has been recently presented [12]. The method addresses the challenge of defining and calculating the functional causal contributions of elements of a biological system. This paper presents the first study applying MSA to the analysis of gene knockout data. The MSA identifies the importance of genes in the Rad6 DNA repair pathway of the yeast S. cerevisiae, quantifying their contributions and characterizing their functional interactions. Incorporating additional biological knowledge, a new functional description of the Rad6 pathway is provided, predicting the existence of additional DNA polymerase and RFC-like complexes. The MSA is the first method for rigorously analyzing multi-knockout experiments, which are likely to soon become a standard and necessary tool for analyzing complex biological systems.

High throughput protein interaction data sets have proven to be notoriously noisy. Although it is possible to focus on interactions with higher reliability by using only those that are backed up by two or more lines of evidence, this approach invariably throws out the majority of available data. A more optimal use could be achieved by incorporating the probabilities associated with all available interactions into the analysis. We present a novel method for estimating error rates associated with specific protein interaction data sets, as well as with individual interactions given the data sets in which they appear. As a bonus, we also get an estimate for the total number of protein interactions in yeast. Certain types of false positive results can be identified and removed, resulting in a significant improvement in quality of the data set. For co-purification data sets, we show how we can reach a tradeoff between the "spoke" and "matrix" representation of interactions within co-purified groups of proteins to achieve an optimal false positive error rate.

Proteases play a fundamental role in the control of intra- and extracellular processes by binding and cleaving specific amino acid sequences. Identifying these targets is extremely challenging. Current computational attempts to predict cleavage sites are limited, representing these amino acid sequences as patterns or frequency matrices. Here we present PoPS, a publicly accessible bioinformatics tool (http://pops.csse.monash.edu.au/) which provides a novel method for building computational models of protease specificity that, while still being based on these amino acid sequences, can be built from any experimental data or expert knowledge available to the user. PoPS specificity models can be used to predict and rank likely cleavages within a single substrate, and within entire proteomes. Other factors, such as the secondary or tertiary structure of the substrate, can be used to screen unlikely sites. Furthermore, the tool also provides facilities to infer, compare and test models, and to store them in a publicly accessible database.

DNA-protein alignment algorithms can be used to discover coding sequences in a genomic sequence, if the corresponding protein derivatives are known. They can also be used to identify potential coding sequences of a newly sequenced genome, by using proteins from related species. Previously known algorithms either solve a simplified formulation, or sacrifice optimality to achieve practical implementation. In this paper, we present a comprehensive formulation of the DNA-protein alignment problem, and an algorithm to compute the optimal alignment in O(mn) time using only four tables of size (m + 1) x (n + 1), where m and n are the lengths of the DNA and protein sequences, respectively. We also developed a Protein and DNA Alignment program PanDA that implements the proposed solution. Experimental results indicate that our algorithm produces high quality alignments.

Gene expression profiles with clinical outcome data enable monitoring of disease progression and prediction of patient survival at the molecular level. We present a new computational method for outcome prediction. Our idea is to use an informative subset of original training samples. This subset consists of only short-term survivors who died within a short period and long-term survivors who were still alive after a long follow-up time. These extreme training samples yield a clear platform to identify genes whose expression is related to survival. To find relevant genes, we combine two feature selection methods -- entropy measure and Wilcoxon rank sum test -- so that a set of sharp discriminating features are identified. The selected training samples and genes are then integrated by a support vector machine to build a prediction model, by which each validation sample is assigned a survival/relapse risk score for drawing Kaplan-Meier survival curves. We apply this method to two data sets: diffuse large-B-cell lymphoma (DLBCL) and primary lung adenocarcinoma. In both cases, patients in high and low risk groups stratified by our risk scores are clearly distinguishable. We also compare our risk scores to some clinical factors, such as International Prognostic Index score for DLBCL analysis and tumor stage information for lung adenocarcinoma. Our results indicate that gene expression profiles combined with carefully chosen learning algorithms can predict patient survival for certain diseases.

It is still an open problem to identify functional relations with o(N . n(k)) time for any domain[2], where N is the number of learning instances, n is the number of genes (or variables) in the Gene Regulatory Network (GRN) models and k is the indegree of the genes. To solve the problem, we introduce a novel algorithm, DFL (Discrete Function Learning), for reconstructing qualitative models of GRNs from gene expression data in this paper. We analyze its complexity of O(k . N . n(2)) on the average and its data requirements. We also perform experiments on both synthetic and Cho et al. [7] yeast cell cycle gene expression data to validate the efficiency and prediction performance of the DFL algorithm. The experiments of synthetic Boolean networks show that the DFL algorithm is more efficient than current algorithms without loss of prediction performances. The results of yeast cell cycle gene expression data show that the DFL algorithm can identify biologically significant models with reasonable accuracy, sensitivity and high precision with respect to the literature evidences. We further introduce a method called epsilon function to deal with noises in data sets. The experimental results show that the epsilon function method is a good supplement to the DFL algorithm.

Feature selection and weighting are central problems in pattern recognition and instance-based learning. In this work, we discuss the challenges of constructing and weighting features to recognize 3D patterns of electron density to determine protein structures. We present SLIDER, a feature-weighting algorithm that adjusts weights iteratively such that patterns that match query instances are better ranked than mismatching ones. Moreover, SLIDER makes judicious choices of weight values to be considered in each iteration, by examining specific weights at which matching and mismatching patterns switch as nearest neighbors to query instances. This approach reduces the space of weight vectors to be searched. We make the following two main observations: (1) SLIDER efficiently generates weights that contribute significantly in the retrieval of matching electron density patterns; (2) the optimum weight vector is sensitive to the distance metric i.e. feature relevance can be, to a certain extent, sensitive to the underlying metric used to compare patterns.

We introduce novel profile-based string kernels for use with support vector machines (SVMs) for the problems of protein classification and remote homology detection. These kernels use probabilistic profiles, such as those produced by the PSI-BLAST algorithm, to define position-dependent mutation neighborhoods along protein sequences for inexact matching of k-length subsequences ("k-mers") in the data. By use of an efficient data structure, the kernels are fast to compute once the profiles have been obtained. For example, the time needed to run PSI-BLAST in order to build the pro- files is significantly longer than both the kernel computation time and the SVM training time. We present remote homology detection experiments based on the SCOP database where we show that profile-based string kernels used with SVM classifiers strongly outperform all recently presented supervised SVM methods. We also show how we can use the learned SVM classifier to extract "discriminative sequence motifs" -- short regions of the original profile that contribute almost all the weight of the SVM classification score -- and show that these discriminative motifs correspond to meaningful structural features in the protein data. The use of PSI-BLAST profiles can be seen as a semi-supervised learning technique, since PSI-BLAST leverages unlabeled data from a large sequence database to build more informative profiles. Recently presented "cluster kernels" give general semi-supervised methods for improving SVM protein classification performance. We show that our profile kernel results are comparable to cluster kernels while providing much better scalability to large datasets.

Distinct RNA structures are frequently involved in a wide-range of functions in various biological mechanisms. The three dimensional RNA structures solved by X-ray crystallography and various well-established RNA phylogenetic structures indicate that functional RNAs have characteristic RNA structural motifs represented by specific combinations of base pairings and conserved nucleotides in the loop region. Discovery of well-ordered RNA structures and their homologues in genome-wide searches will enhance our ability to detect the RNA structural motifs and help us to highlight their association with functional and regulatory RNA elements. We present here a novel computer algorithm, HomoStRscan, that takes a single RNA sequence with its secondary structure to search for homologous-RNAs in complete genomes. This novel algorithm completely differs from other currently used search algorithms of homologous structures or structural motifs. For an arbitrary segment (or window) given in the target sequence, that has similar size to the query sequence, HomoStRscan finds the most similar structure to the input query structure and computes the maximal similarity score (MSS) between the two structures. The homologousRNA structures are then statistically inferred from the MSS distribution computed in the target genome. The method provides a flexible, robust and fine search tool for any homologous structural RNAs.