Computational systems bioinformatics. Computational Systems Bioinformatics Conference最新文献

Methods that can screen large databases to retrieve a structurally diverse set of compounds with desirable bioactivity properties are critical in the drug discovery and development process. This paper presents a set of such methods, which are designed to find compounds that are structurally different to a certain query compound while retaining its bioactivity properties (scaffold hops). These methods utilize various indirect ways of measuring the similarity between the query and a compound that take into account additional information beyond their structure-based similarities. Two sets of techniques are presented that capture these indirect similarities using approaches based on automatic relevance feedback and on analyzing the similarity network formed by the query and the database compounds. Experimental evaluation shows that many of these methods substantially outperform previously developed approaches both in terms of their ability to identify structurally diverse active compounds as well as active compounds in general.

In recent years, biclique methods have been proposed to construct phylogenetic trees. One of the key steps of these methods is to find complete sub-matrices (without missing entries) from a species-genes data matrix. To enumerate all complete sub-matrices, (17) described an exact algorithm, whose running time is exponential. Furthermore, it generates a large number of complete sub-matrices, many of which may not be used for tree reconstruction. Further investigating and understanding the characteristics of species-genes data may be helpful for discovering complete sub-matrices. Therefore, in this paper, we focus on quantitatively studying and understanding the characteristics of species-genes data, which can be used to guide new algorithm design for efficient phylogenetic inference. In this paper, a mathematical model is constructed to simulate the real species-genes data. The results indicate that sequence-availability probability distributions follow power law, which leads to the skewness and sparseness of the real species-genes data. Moreover, a special structure, called "ladder structure", is discovered in the real species-genes data. This ladder structure is used to identify complete sub-matrices, and more importantly, to reveal overlapping relationships among complete sub-matrices. To discover the distinct ladder structure in real species-genes data, we propose an efficient evolutionary dynamical system, called "generalized replicator dynamics". Two species-genes data sets from green plants are used to illustrate the effectiveness of our model. Empirical study has shown that our model is effective and efficient in understanding species-genes data for phylogenetic inference.

The ability to identify gene mentions in text and normalize them to the proper unique identifiers is crucial for "down-stream" text mining applications in bioinformatics. We have developed a rule-based algorithm that divides the normalization task into two steps. The first step includes pattern matching for gene symbols and an approximate term searching technique for gene names. Next, the algorithm measures several features based on morphological, statistical, and contextual information to estimate the level of confidence that the correct identifier is selected for a potential mention. Uniqueness, inverse distance, and coverage are three novel features we quantified. The algorithm was evaluated against the BioCreAtIvE datasets. The feature weights were tuned by the Nealder-Mead simplex method. An F-score of .7622 and an AUC (area under the recall-precision curve) of .7461 were achieved on the test data using the set of weights optimized to the training data.

To understand the regulation of the gene expression, the identification of transcription start sites (TSSs) is a primary and important step. With the aim to improve the computational prediction accuracy, we focus on the most challenging task, i.e., to identify the TSSs within 50 bp in non-CpG related promoter regions. Due to the diversity of non-CpG related promoters, a large number of features are extracted. Effective feature selection can minimize the noise, improve the prediction accuracy, and also to discover biologically meaningful intrinsic properties. In this paper, a newly proposed multi-objective simulated annealing based optimization method, Archive Multi-Objective Simulated Annealing (AMOSA), is integrated with Linear Discriminant Analysis (LDA) to yield a combined feature selection and classification system. This system is found to be comparable to, often better than, several existing methods in terms of different quantitative performance measures.

Reconciliation is the process of resolving disagreement between gene and species trees, by invoking gene duplications and losses to explain topological incongruence. The resulting inferred duplication histories are a valuable source of information for a broad range of biological applications, including ortholog identification, estimating gene duplication times, and rooting and correcting gene trees. Reconciliation for binary trees is a tractable and well studied problem. However, a striking proportion of species trees are non-binary. For example, 64% of branch points in the NCBI taxonomy have three or more children. When applied to non-binary species trees, current algorithms overestimate the number of duplications because they cannot distinguish between duplication and deep coalescence. We present the first formal algorithm for reconciling binary gene trees with non-binary species trees under a duplication-loss parsimony model. Using a space efficient mapping from gene to species tree, our algorithm infers the minimum number of duplications and losses in O(|V(G)| . (k(S) + h(S))) time, where V(G) is the number of nodes in the gene tree, h(S) is the height of the species tree and k(S) is the width of its largest multifurcation. We also present a dynamic programming algorithm for a combined loss model, in which losses in sibling species may be represented as a single loss in the common ancestor. Our algorithms have been implemented in NOTUNG, a robust, production quality tree-fitting program, which provides a graphical user interface for exploratory analysis and also supports automated, high-throughput analysis of large data sets.