Catálogo de publicaciones - libros

Hearing disorders affect over 10% of the population and this ratio is dramatically increasing with age. Development of appropriate therapeutic approaches requires understanding of the auditory system, which remains largely incomplete. We have identified hearing-specific genes and pathways by mapping over 15000 cochlear expressed sequence tags (ESTs) to the human genome (NCBI Build 35) and comparing it to other EST clusters (Unigene Build 183). A number of novel potentially cochlear-specific genes discovered in this work are currently being verified by experimental studies. The software tool developed for this task is based on a fast bidirectional multiple pattern search algorithm. Patterns used for scoring and selection of loci include EST subsequences, cloning-process identifiers, and genomic and external contamination determinants. Comparison of our results with other programs and available annotations shows that the software developed provides potentially the fastest, yet reliable mapping of ESTs.

Non-coding RNAs (ncRNAs) are functional RNA molecules that do not code for proteins. Classic examples include ribosomal and transfer RNAs, but dramatic discoveries in the last few years have greatly expanded both the number of known ncRNAs and the breadth of their biological roles [1]. In short, ncRNAs are much more biologically significant than previously realized.

The computational problems associated with discovery and characterization of ncRNAs are quite different from, and arguably more difficult than, comparable tasks for protein-coding genes [2]. A key element of this difference is the importance of secondary structure in most ncRNAs. RNA secondary structure prediction is a well-studied problem, and useful tools exist, but they are certainly not perfect. It is generally accepted that the best evidence for stable secondary structure in biologically relevant RNAs is to identify diverged examples exhibiting compensatory base-pair changes that would preserve putative structural elements. Unfortunately, such compensatory mutations interfere with the ability of standard sequence search and alignment tools (e.g., BLAST, ClustalW) to find and align homologs.

The application of new information and biotechnologies to infectious disease research provides an opportunity to design, develop and deploy a comprehensive cyberinfrastructure for life sciences. The application of integrative approaches including theory, wet experimentation, modeling and simulation and the leveraging of a strong comparative, evolutionary framework has spawned pathosystems biology. I will show examples of how cyberinfrastructure is being developed and used to support pathosystems biology.

The recently developed technology of genomic tiling microarrays, which can be used for genome annotation, has required the development of new methodologies [Royce et.al] for their design and analysis. Genomic tiling arrays use PCR amplicons or short oligonucleotide probes to the non-repetitive DNA sequence of a genome in an unbiased fashion for the purposes of detecting novel genomic features. Specifically, they can be used for the identification of novel transcripts, distinguishing between different splice isoforms and for finding transcription factor binding sites using Chromatin-Immunoprecipitation on chip experiments (ChIP-chip).

Reactions within the cell should satisfy the law of mass conservation and the second law of thermodynamics. Networks of reactions violating any of these laws are unphysical and cannot occur in nature. In this paper we describe a technique that perturbs an unfeasible network to produce a metabolic network that satisfies the two fundamental laws. This algorithm has been applied to study the metabolic pathways of .

Many cellular functions are carried out in compartments of the cell. The cellular localization of a protein is thus related to its function identification. This paper investigates the use of two Machine Learning techniques, Support Vector Machines (SVMs) and Decision Trees (DTs), in the protein cellular localization prediction problem. Since the given task has multiple classes and SVMs are originally designed for the solution of two class problems, several strategies for multiclass SVMs extension were investigated, including one proposed by the authors.

One-class classification techniques are able to, based only on examples of a normal profile, induce a classifier that is capable of identifying novel classes or profile changes. However, the performance of different novelty detection approaches may depend on the domain considered. This paper applies combined one-class classifiers to detect novelty in gene expression data. Results indicate that the robustness of the classification is increased with this combined approach.

This work investigates the behavior of two different clustering algorithms, with two proximity measures, in terms of the contents of the partitions obtained with them. An analysis of how the classes are separated by these algorithms, as different numbers of clusters are generated, is also presented. A discussion on the use of these information in the identification of special cases for further analysis by biologists is presented.

One of the most important goals of bioinformatics is the ability to identify genes in uncharacterized DNA sequences. Improved promoter prediction methods can be one step towards developing more reliable gene prediction methods. In this paper, we present an empirical comparison of machine learning techniques such as Naive Bayes, Decision Trees, Support Vector Machines and Neural Networks to the task of predicting promoters. In order to do so, we first built a data set of promoter and nonpromoter sequences for this organism.

Gene expression on prokaryotes initiates when the RNA-polymerase enzyme interacts with DNA regions called promoters, where are located the main regulatory elements of the transcription process. Despite the improvement of techniques for molecular biology analysis, characterizing and identifying promoters is a complex task. approaches are used to recognize theses regions. Nevertheless, they confront the absence of a large set of promoters to identify conserved patterns among the species. Hence, a methodology able to predict them on any genome is a challenge. This work proposes a methodology based on Hidden Markov Models (HMMs), Decision Threshold Estimation and Discrimination Analysis. For three investigated prokaryotic species, the mainly results are: a reduction in 44.96% of recognition error rate compared with previous works on , an accuracy of 95% on recognition and 78% on prediction for . However, it was found a large number of false positives on