Abstract not found

Motivation: Bayesian network methods have shown promise in gene regulatory network reconstruction because of their capability of capturing causal relationships between genes and handling data with noises found in biological experiments. The problem of learning network structures, however, is NP hard. Consequently, heuristic methods such as hill climbing are used for structure learning. For networks of a moderate size, hill climbing methods are not computationally efficient. Furthermore, relatively low accuracy of the learned structures may be observed. The purpose of this paper is to present a novel structure learning method for gene network discovery.. Results: In this paper, we present a novel structure learning method to reconstruct the underlying gene networks from the observational gene expression data. Unlike hill climbing approaches, the proposed method first constructs an undirected network based on mutual information between two nodes and then split the structure into substructures. The directional orientations for the edges that connect two nodes are then obtained by optimizing a scoring function for each substructure. Our method is evaluated using two benchmark network datasets with known structures. The results show that the proposed method can identify networks that are close to the optimal structures. It outperforms hill climbing methods in terms of both computation time and predicted structure accuracy. We also apply the method to gene expression data measured during the yeast cycle and show the effectiveness of the proposed method for network reconstruction.

Abstract. Recent experimental advances facilitate the collection of time series data that indicate which genes in a cell are expressed. This paper proposes an efficient method to generate the genetic regulatory network inferred from time series data. Our method first encodes the data into levels. Next, it determines the set of potential parents for each gene based upon the probability of the gene’s expression increasing. After a subset of potential parents are selected, it determines if any genes in this set may have a combined effect. Finally, the potential sets of parents are competed against each other to determine the final set of parents. The result is a directed graph representation of the genetic network’s repression and activation connections. Our results on synthetic data generated from models for several genetic networks with tight feedback are promising. 1

Abstract. Inferring transcriptional regulatory networks from gene-expression data remains a challenging problem, in part because of the noisy nature of the data and the lack of strong network models. Time-series expression data have shown promise and recent work by Babu on the evolution of regulatory networks in E. coli and S. cerevisiae opened another avenue of investigation. In this paper we take the evolutionary approach one step further, by developing ML-based refinement algorithms that take advantage of established phylogenetic relationships among a group of related organisms and of a simple evolutionary model for regulatory networks to improve the inference of these networks for these organisms from expression data gathered under similar conditions. We use simulations with different methods for generating gene-expression data, different phylogenies, and different evolutionary rates, and use different network inference algorithms, to study the performance of our algorithmic boosters. The results of simulations (including various tests to exclude confounding factors) demonstrate clear and significant improvements (in both specificity and sensitivity) on the performance of current inference algorithms. Thus geneexpression studies across a range of related organisms could yield significantly more accurate regulatory networks than single-organism studies. 1

Abstract. Learning the structure of a gene regulatory network from time-series gene expression data is a significant challenge. Most approaches proposed in the literature to date attempt to predict the regulators of each target gene individually, but fail to share regulatory information between related genes. In this paper, we propose a new globally regularized risk minimization approach to address this problem. Our approach first clusters genes according to their time-series expression profiles— identifying related groups of genes. Given a clustering, we then develop a simple technique that exploits the assumption that genes with similar expression patterns are likely to be co-regulated by encouraging the genes in the same group to share common regulators. Our experiments on both synthetic and real gene expression data suggest that our new approach is more effective at identifying important transcription factor based regulatory mechanisms than the standard independent approach and a prototype based approach. 1

Abstract. Computational inference of transcriptional regulatory networks remains a challenging problem, in part due to the lack of strong network models. In this paper we present evolutionary approaches to improve the inference of regulatory networks for a family of organisms by developing an evolutionary model for these networks and taking advantage of established phylogenetic relationships among these organisms. In previous work, we used a simple evolutionary model for regulatory networks and provided extensive simulation results showing that phylogenetic information, combined with such a model, could be used to gain significant improvements on the performance of current inference algorithms. In this paper, we extend the evolutionary model so as to take into account gene duplications and losses, which are viewed as major drivers in the evolution of regulatory networks. We show how to adapt our evolutionary approach to this new model and provide detailed simulation results, which show significant improvement on the reference network inference algorithms. We also provide results on biological data (cis-regulatory modules for 12 species of Drosophila), confirming our simulation results. 1

A dynamic Bayesian network (DBN) is a probabilistic network that models interdependent entities that change over time. Given example sequences of multivariate data, we use a genetic algorithm to synthesize a network structure that models the causal relationships that explain the sequence. We use a multi-objective evaluation strategy with a genetic algorithm. The multi-objective criteria are a network’s probabilistic score and structural complexity score. Our use of Pareto ranking is ideal for this application, because it naturally balances the effect of the likelihood and structural simplicity terms used in the BIC network evaluation heuristic. We use a simple structural scoring formula, which tries to keep the number of links in the network approximately equivalent to the number of variables. We also use a simple representation that favours sparsely connected networks similar in structure to those modeling biological phenomenon. Our experiments show promising results when evolving networks ranging from 10 to 30 variables, using a maximal connectivity of between 3 and 4 parents per node. The results from the multi-objective GA were superior to those obtained with a single objective GA. Key words: dynamic Bayesian networks, multi-objective optimization, genetic algorithms

The understanding of molecular cell biology requires insight into the structure and dynamics of networks that are made up of thousands of interacting molecules of DNA, RNA, proteins, metabolites, and other components. One of the central goals of systems biology is the unraveling of the as yet poorly

In our group we have often encountered the need to evaluate the efficacy of our reverse engineering algorithms. Our evaluation attempts can be divided into two categories: 1. evaluations using simulation

I would like to express my sincere gratitude to my supervisors, Dr. Sung Wing-Kin, Dr. Mao Pei-Lin and Dr. Liu Bing, for providing me with the wonderful opportunity to pursue my PhD degree. I am grateful to them for their continuous encouragement, support and guidance throughout of years of my study. I am thankful to the graduate supervisory committee overseeing my work, Dr. Tung Kum Hoe and Dr. Lee Wee Sun for their constructive suggestions and critical comments. Special thanks go to Dr. Wu Ping and Dr. Ankush Mittal for their guidance as well as helpful suggestions. Madam Leong Yoke Yee is also highly appreciated for helping me refine the thesis. I thank all past and present members of the computational biology lab for their idea sharing. The wonderful time we have spent together in NUS will be in my mind forever. My heartfelt appreciation goes to my beloved parents for their constant support and encouragement, without whom this would have remained but a dream. Finally, my deepest gratitude goes to my wife for her unconditional love, understanding and warm support through the years. ii