Takeyuki Tamura

Subcellular location prediction of proteins using support vector machines with alignment of block sequences utilizing amino acid composition

BackgroundSubcellular location prediction of proteins is an important and well-studied problem in bioinformatics. This is a problem of predicting which part in a cell a given protein is transported to, where an amino acid sequence of the protein is given as an input. This problem is becoming more important since information on subcellular location is helpful for annotation of proteins and genes and the number of complete genomes is rapidly increasing. Since existing predictors are based on various heuristics, it is important to develop a simple method with high prediction accuracies.ResultsIn this paper, we propose a novel and general predicting method by combining techniques for sequence alignment and feature vectors based on amino acid composition. We implemented this method with support vector machines on plant data sets extracted from the TargetP database. Through fivefold cross validation tests, the obtained overall accuracies and average MCC were 0.9096 and 0.8655 respectively. We also applied our method to other datasets including that of WoLF PSORT.ConclusionAlthough there is a predictor which uses the information of gene ontology and yields higher accuracy than ours, our accuracies are higher than existing predictors which use only sequence information. Since such information as gene ontology can be obtained only for known proteins, our predictor is considered to be useful for subcellular location prediction of newly-discovered proteins. Furthermore, the idea of combination of alignment and amino acid frequency is novel and general so that it may be applied to other problems in bioinformatics. Our method for plant is also implemented as a web-system and available on http://sunflower.kuicr.kyoto-u.ac.jp/~tamura/slpfa.html.

COMPOUND ANALYSIS VIA GRAPH KERNELS INCORPORATING CHIRALITY

High accuracy is paramount when predicting biochemical characteristics using Quantitative Structural-Property Relationships (QSPRs). Although existing graph-theoretic kernel methods combined with machine learning techniques are efficient for QSPR model construction, they cannot distinguish topologically identical chiral compounds which often exhibit different biological characteristics. In this paper, we propose a new method that extends the recently developed tree pattern graph kernel to accommodate stereoisomers. We show that Support Vector Regression (SVR) with a chiral graph kernel is useful for target property prediction by demonstrating its application to a set of human vitamin D receptor ligands currently under consideration for their potential anti-cancer effects.

Finding a Periodic Attractor of a Boolean Network

In this paper, we study the problem of finding a periodic attractor of a Boolean network (BN), which arises in computational systems biology and is known to be NP-hard. Since a general case is quite hard to solve, we consider special but biologically important subclasses of BNs. For finding an attractor of period 2 of a BN consisting of n OR functions of positive literals, we present a polynomial time algorithm. For finding an attractor of period 2 of a BN consisting of n AND/OR functions of literals, we present an O(1.985n) time algorithm. For finding an attractor of a fixed period of a BN consisting of n nested canalyzing functions and having constant treewidth w, we present an O(n2p(w+1)poly(n)) time algorithm.

Completing networks using observed data

This paper studies problems of completing a given Boolean network (Boolean circuit) so that the input/output behavior is consistent with given examples, where we only consider acyclic networks. These problems arise in the study of inference of signaling networks using reporter proteins. We prove that these problems are NP-complete in general and a basic version remains NP-complete even for tree structured networks. On the other hand, we show that these problems can be solved in polynomial time for partial k-trees of bounded (constant) indegree if a logarithmic number of examples are given.

Determining a Singleton Attractor of a Boolean Network with Nested Canalyzing Functions

In this article, we study the problem of finding a singleton attractor for several biologically important subclasses of Boolean networks (BNs). The problem of finding a singleton attractor in a BN is known to be NP-hard in general. For BNs consisting of n nested canalyzing functions, we present an O(1.799(n)) time algorithm. The core part of this development is an O(min(2(k/2) · 2(m/2), 2(k)) · poly(k, m)) time algorithm for the satisfiability problem for m nested canalyzing functions over k variables. For BNs consisting of chain functions, a subclass of nested canalyzing functions, we present an O(1.619(n)) time algorithm and show that the problem remains NP-hard, even though the satisfiability problem for m chain functions over k variables is solvable in polynomial time. Finally, we present an o(2(n)) time algorithm for bounded degree BNs consisting of canalyzing functions.

Determining a singleton attractor of an AND/OR Boolean network in O (1.587n) time

The Boolean network (BN) is a discrete model of gene regulatory networks [5]. Each node in this network corresponds to a gene, and takes on a value of 1 or 0, meaning that the gene is or is not expressed. The value of a node at a given time instant is determined according to a regulation rule that is a Boolean function of the values of the predecessors of the node at the previous time, or their negations. The values of nodes change synchronously. We focus here on AND/OR Boolean networks, in which the regulation rule assigned to each node is restricted to be either a conjunction or a disjunction of literals. An important characteristic of any BN is the existence of an attractor, whether it is a singleton attractor, i.e. a stable state, or a cyclic attractor, i.e. a state that repeats periodically. Here the state of a network at a given time instant is the set of its node values. Unfortunately, the problem of detection of a singleton attractor (or an attractor of the shortest

A clique-based method for the edit distance between unordered trees and its application to analysis of glycan structures.

BackgroundMeasuring similarities between tree structured data is important for analysis of RNA secondary structures, phylogenetic trees, glycan structures, and vascular trees. The edit distance is one of the most widely used measures for comparison of tree structured data. However, it is known that computation of the edit distance for rooted unordered trees is NP-hard. Furthermore, there is almost no available software tool that can compute the exact edit distance for unordered trees.ResultsIn this paper, we present a practical method for computing the edit distance between rooted unordered trees. In this method, the edit distance problem for unordered trees is transformed into the maximum clique problem and then efficient solvers for the maximum clique problem are applied. We applied the proposed method to similar structure search for glycan structures. The result suggests that our proposed method can efficiently compute the edit distance for moderate size unordered trees. It also suggests that the proposed method has the accuracy comparative to those by the edit distance for ordered trees and by an existing method for glycan search.ConclusionsThe proposed method is simple but useful for computation of the edit distance between unordered trees. The object code is available upon request.

Finding Minimum Reaction Cuts of Metabolic Networks Under a Boolean Model Using Integer Programming and Feedback Vertex Sets

In this paper, the authors consider the problem of, given a metabolic network, a set of source compounds and a set of target compounds, finding a minimum size reaction cut, where a Boolean model is used as a model of metabolic networks. The problem has potential applications to measurement of structural robustness of metabolic networks and detection of drug targets. They develop an integer programming-based method for this optimization problem. In order to cope with cycles and reversible reactions, they further develop a novel integer programming (IP) formalization method using a feedback vertex set (FVS). When applied to an E. coli metabolic network consisting of Glycolysis/Glyconeogenesis, Citrate cycle and Pentose phosphate pathway obtained from KEGG database, the FVS-based method can find an optimal set of reactions to be inactivated much faster than a naive IP-based method and several times faster than a flux balance-based method. The authors also confirm that our proposed method works even for large networks and discuss the biological meaning of our results.

A Clique-Based Method Using Dynamic Programming for Computing Edit Distance Between Unordered Trees

Many kinds of tree-structured data, such as RNA secondary structures, have become available due to the progress of techniques in the field of molecular biology. To analyze the tree-structured data, various measures for computing the similarity between them have been developed and applied. Among them, tree edit distance is one of the most widely used measures. However, the tree edit distance problem for unordered trees is NP-hard. Therefore, it is required to develop efficient algorithms for the problem. Recently, a practical method called clique-based algorithm has been proposed, but it is not fast for large trees. This article presents an improved clique-based method for the tree edit distance problem for unordered trees. The improved method is obtained by introducing a dynamic programming scheme and heuristic techniques to the previous clique-based method. To evaluate the efficiency of the improved method, we applied the method to comparison of real tree structured data such as glycan structures. For large tree-structures, the improved method is much faster than the previous method. In particular, for hard instances, the improved method achieved more than 100 times speed-up.

Network Completion Using Dynamic Programming and Least-Squares Fitting

We consider the problem of network completion, which is to make the minimum amount of modifications to a given network so that the resulting network is most consistent with the observed data. We employ here a certain type of differential equations as gene regulation rules in a genetic network, gene expression time series data as observed data, and deletions and additions of edges as basic modification operations. In addition, we assume that the numbers of deleted and added edges are specified. For this problem, we present a novel method using dynamic programming and least-squares fitting and show that it outputs a network with the minimum sum squared error in polynomial time if the maximum indegree of the network is bounded by a constant. We also perform computational experiments using both artificially generated and real gene expression time series data.