Catálogo de publicaciones - libros

We developed a methodology to predict myotoxicity and neurotoxicity of proteins of the family of Phospholipases A2 (PLA2) from sequence data. Combining two bioinformatics tools, MEME and HMMER, it was possible to detect conserved motifs and represent them as Hidden Markov Models (HMMs). In ten-fold cross validation testing we have determined the efficacy of each motif on prediction of PLA2 function. We selected motifs whose efficacy in predict function were above 60 % at the Minimum Error Point (MEP), the score in which there are fewest both false positives and false negatives. Combining HMMs of the best motifs for each function, we have achieved a mean efficacy of 98 ± 4 % on prediction of myotoxic function and 77.4 ± 4.8% on prediction of neurotoxicity. We have used the results of this work to build a web tool (available at www.cbiot.ufrgs.br/bioinfo/ phospholipase) to classify PLA2s of unknown function regarding myotoxic or neurotoxic activity.

, a trematode parasite, is the major causative agent of Schistomiasis, a public health problem in South America and Africa. Recently, as a result of two separate efforts, the transcriptome of [1] and [2] were published. Schistosomes possess distinct and differentiated organs and have evolved to adapt to parasitism. Availability of transcriptome data has raised a number of issues regarding the parasite’s cell biology and signaling pathways, as recently discussed in a review [3].

In order to select gene markers among differentially expressed transcripts identified from tumoral prostate, we have applied a filter and Significance Analysis of Microarrays (SAM) as the feature selection method on a previously normalized dataset of DNA microarray experiments reported by Reis et al., 2004 (Oncogene 23:6684-6692). Twenty seven samples with different degrees of tumor differentiation (Gleason scores) were analyzed. SAM was run using either two-class, unpaired data analysis with Gleason 5-6 and Gleason 9-10 samples, or multiclass response analysis with an additional category of Gleason 7-8. Both strategies revealed a promising set of transcripts associated with the degree of differentiation of prostate tumors.

Trimming procedures are an important part of the sequence analysis pipeline in an EST Sequencing Project. In general, trimming is done in several phases, each one detecting and removing some kind of undesirable artifact, such as low quality sequence, vectors or adapters, and contamination. However, this strategy often results in a phase being unable to recognize its target because part of it was removed during a previous phase. To remedy this drawback, we propose a new strategy, where each phase detects but does not remove its target, leaving this decision to a post processing step occurring after all phases. Our tests show that this strategy can significantly improve the detection of artifacts.

Landau and Vishkin developed an () algorithm for the approximate string matching problem, where is the maximum number of admissible errors and the length of the text. This algorithm uses suffix trees for an (1) running time computation of the longest common extensions between strings. We present a variation of this algorithm which uses suffix arrays for computing the longest common extensions.

This paper describes BioPAUÁ Project, a new portal for Molecular Dynamics (MD) simulations over a computational grid environment. It integrates MD simulations and analyses tools with grid technologies to provide support to biomolecular experiments. The objective of BioPAUÁ Project is to offer a tool, as well the facility, for researches working in several important fields ( bioinformatics, structural biology, biochemistry, medicinal chemistry, biopharmacology). At first, the possible user does not need any special skill in MD simulations, however, advanced ones are also well contemplated. The project methodology is based on MYGRID middleware and uses GROMACS package in order to run simulations. This work is developed by LNCC/MCT, with IBCCF/UFRJ collaboration, and supported by HP Brazil R&D.

Methyl-CpG-binding protein 2 (MeCP2) belongs to the DNA-binding protein family that selectively binds to DNA methylated CpG-islands. MeCP2 acts like a transcriptional repressor, that contains a N-terminal methylated DNA-binding domain (MBD), and a C-terminal transcriptional repression domain (TRD). Mutations in MECP2 gene have been associated to Rett Syndrome – a neurological disorder linked to X-chromossome, and one of the most common causes of physical and intellectual dysfunction in females. The calculation of MeCP2 MDB had been solved, but the effects of the mutations on the protein’s structure and, consequently, functions have not been analyzed. Databases, systems, tools, and, more recently, protein structure motifs databases available on Internet make it possible to predict ab initio protein structure quickly. This extended abstract looks at the the use of these tools to analyze the effects of MeCP2’s mutations, which cause Rett syndrome, in the original protein structure.

We used Artificial Neural Network for protein loop classification based on the amino acid sequence alone. A new algorithm recently proposed, the Hidden Layer Learning Vector Quantization (HLVQ) was used and its accuracy compared with traditional Multilayer Preceptrons (MLP). The HLVQ algorithm achieved superior accuracy correctly classifying most loops.

Protein structure visualization is crucial for understanding its function inside the cell. Each year, laboratories around the world deposit protein structures on a central database for further analysis and research. The result is a large amount of structures being deposited (approximately 31,000 in may 2005). Visualization is a very powerful tool to help in the analysis, aiding data understanding and interpretation. The present work suggests an architecture to help the rapid construction of visual biomolecular software, specifically designed to be simple, modular and scalable. The architecture, called , employs high quality open-source libraries offering simple data structures and customizable options. The architecture can be used to start a new visual software project to visualize and represent individual protein structures, as well as multiple conformations from molecular dynamics simulation trajectories.

Automatic annotation tools are becoming popular since the biologists and curators of databases cannot cope with the volume of sequences to be annotated manually. One way to automate the annotation is to use techniques of symbolic machine learning to derive rules to guide this annotation. However, the training instances tend to have too many attributes, turning the machine learning process difficult and time consuming.