Search

The inference of a network structure from microarray data providing dynamical information about the underlying Gene regulatory network is an important and still outstanding problem. Recently, a causal modeling approach was presented in our publications to recover the structure of this network. However, issues like spurious arcs and time delay were not dealt with previously. The graph-theoretical measure d-separation provides criteria to recover the network structure edge-by-edge by calculating the partial correlation. However, the estimation of partial correlations from small sample sizes is a practical problem. As our approach attempts to find networks that closely match the observed partial correlation constraints in the data, main aim to this paper is to attempt to maximize the scoring metric used. In this paper, we formulate a framework for path analysis as a post processing step after learning gene regulatory network using causal modeling. The approach is tested with both artificial and real gene regulatory network scenario and the structure recovered after post processing better fits the data. [ABSTRACT FROM AUTHOR]

Human endogenous retroviruses (HERVs) are remnants of ancient retrovirus infections and now reside within the human DNA. Recently HERV expression has been detected in both normal and diseased tissues. However, the patterns of expression of individual HERV sequences are mostly unknown. In this work we use a generative mixture model, based on hidden Markov models, for estimating the activities of individual HERV sequences from databases of expressed sequences. We determine the relative activities of sixty HERVs from the HML2 group in five human tissues, i.e. we estimate the expression profile of each HERV. This allows us to gain insight into HERV function. [ABSTRACT FROM AUTHOR]

This paper introduces a novel gene selection method incorporating mutual information in the support vector machine recursive feature elimination (SVM-RFE). We incorporate an additional term of mutual information based minimum redundancy maximum relevancy criteria along with feature weight calculated by SVM algorithm. We tested proposed method on colon cancer and leukemia cancer gene expression dataset. The results show that the proposed method performs better than the original SVM-RFE method. The selected gene subset has better classification accuracy and better generalization capability. [ABSTRACT FROM AUTHOR]

A method for automatic classification of computed tomography (CT) brain images of different head trauma types is presented in this paper. The method has three major steps: 1. The images are first segmented to find potential hemorrhage regions using ellipse fitting, background removal and wavelet decomposition technique; 2. For each region, features (such as area, major axis length, etc.) are extracted; 3. Each extracted feature is classified using machine learning algorithm; the images are then classified based on its component regions' classification. The automatic medical image classification will be useful in building a content-based medical image retrieval system. [ABSTRACT FROM AUTHOR]

Translation of human intentions into control signals for a computer, so called Brain-Computer Interface (BCI), has been a growing research field during the last years. In this way, classification of mental tasks is under investigation in the BCI society as a basic research. In this paper, a Weighted Distance Nearest Neighbor (WDNN) classifier is presented to improve the classification rate between the left and right imagery tasks in which a weight is assigned to each stored instance. The specified weight of each instance is then used for calculating the distance of a test pattern to that instance. We propose an iterative learning algorithm to specify the weights of training instances such that the error rate of the classifier on training data is minimized. ElectroEncephaloGram (EEG) signals are caught from four familiar subjects with the cue-based BCI. The proposed WDNN classifier is applied to the band power and fractal dimension features, which are extracted from EEG signals to classify mental tasks. Results show that our proposed method performs better in some subjects in comparison with the LDA and SVM, as well-known classifiers in the BCI field. [ABSTRACT FROM AUTHOR]

Diffusion Tensor MR Imaging (DTI) provides non-invasive approach to track white matter (WM) trajectories within human brain in vivo, and thereby facilitates studies of anatomical connectivity between sub-cortical and cortical regions. This paper presents a probabilistic fiber tracking framework, which aims to address the two problems in earlier approaches: first, it does not adopt fractional anisotropy (FA) as the stopping criteria so that the exploration of cortico-cortical connectivity is feasible; secondly, fiber tracking process is regularized so that trajectory with low curvature means high belief of connection between two voxels. [ABSTRACT FROM AUTHOR]

A time-frequency method based on Cohen's class of distribution is proposed for analysis of functional magnetic resonance imaging (fMRI) data and to detect activation in the brain regions. The Rihaczek-Margenau distribution among the various distributions of Cohen's class produces the least amount of cross products and is used here for calculating the spectrum of fMRI time-series. This method also does not suffer from the time and frequency resolution trade-off which is inherent in short-term Fourier transform (STFT). Other than detecting activation, the time-frequency analysis is also capable of providing us with more details about the non-stationarity in fMRI data, which can be used for clustering the data into various brain states. The results of brain activation detection with this techniques are presented here and are compared with other prevalent techniques. [ABSTRACT FROM AUTHOR]

Recent advances in bioimaging have allowed to observe biological phenomena in three dimensions in a precise and automated fashion. However, the analysis of depth-stacks acquired in fluorescence microscopy constitutes a challenging task and motivates the development of robust methods. Automated computational schemes to process 3D multi-cell images from High Content Screening (HCS) experiments are part of the next generation methods for drug discovery. Working toward this goal, we propose a fully automated framework which allows fast segmentation and 3D morphometric analysis of cell nuclei. The method is based on deformable models called Active Meshes, featuring automated initialization, robustness to noise, real-time 3D visualization of the objects during their analysis and precise geometrical shape measurements thanks to a parametric representation of each object. The framework has been tested on a low throughput microscope (classically found in research facilities) and on a fully automated imaging platform (used in screening facilities). We also propose shape descriptors and evaluate their robustness and independence on fluorescent beads and on two cell lines. [ABSTRACT FROM AUTHOR]

Identifying antigenic peptides that bind to Major Histocompatibility Complex (MHC) molecules plays a central role in determining T-cell epitopes suitable as vaccine targets. Prediction of the binding ability of antigenic peptides to MHC class II molecules is more complex that for class I. Class II molecules bind to peptides of different lengths and the core region that interacts with the binding site on the class II MHC molecule is located anywhere within the peptide. Obtaining an alignment for these binding sites is an important first step in determining the binding motif of MHC class II alleles. In this paper, we exploit entropy and evolutionary distance of the key binding positions (anchor positions) of an alignment in determining the best possible alignment for a given set of peptide data. Once an optimal alignment is found, a weight matrix representing the binding motif is estimated. The weight matrix designed is subsequently applied to predict MHC binding peptides. [ABSTRACT FROM AUTHOR]

Peptide binding to HLA molecules is a critical step in induction and regulation of T-cell mediated immune responses. Because of combinatorial complexity of immune responses, systematic studies require combination of computational methods and experimentation. Most of available computational predictions are based on discriminating binders from non-binders based on use of suitable prediction thresholds. We compared four state-of-the-art binding affinity prediction models and found that nonlinear models show better performance than linear models. A comprehensive analysis of HLA binders (A*0101, A*0201, A*0301, A*1101, A*2402, B*0702, B*0801 and B*1501) showed that non-linear predictors predict peptide binding affinity with high accuracy. The analysis of known T-cell epitopes of survivin and known HIV T-cell epitopes showed lack of correlation between binding affinity and immunogenicity of HLA-presented peptides. T-cell epitopes, therefore, can not be directly determined from binding affinities by simple selection of the highest affinity binders. [ABSTRACT FROM AUTHOR]

Comparative genomic analysis between pathogens and the host Homo sapiens has led to identification of novel drug targets. Microbial drug target identification and validation has been the latest trend in pharmacoinformatics. In order to identify a suitable drug target for the pathogen Pseudomonas aeruginosa an in silico comparative analysis of the metabolic pathways between the pathogen and the host Homo sapiens was performed. Detection of bacterial genes that are non-homologous to human genes, and are essential for the survival of the pathogen represents a promising means of identifying novel drug targets. Metabolic pathways for the pathogen and H.sapiens were obtained from the metabolic pathway database KEGG and were compared to identify unique pathways present only in the pathogen and absent in the host. We identified 361 enzymes from both unique and common pathways between the pathogen and the host of which 50 belong to the 12 unique pathways. Enzymes from both genomes were subject to a BLASTp search and sequences homologous to human were removed as non essential. P.aeruginosa targets without human homologs were identified when the e-value threshold was set as 10− 2. Of the 214 targets that had no hits only 30 targets belong to unique pathways. These 30 targets were then compared with the list of candidate essential genes identified by mutagenesis. Only 8 targets matched with the essential genes list and these were considered as potential drug targets. We have built homology model for the four target genes lpxC, kdsA, kdsB and waaG using MODELLER software. This approach enables rapid potential drug target identification, thereby greatly facilitating the search for new antibiotics. [ABSTRACT FROM AUTHOR]

Sequential classification task is typical in medical diagnosis, when the investigations of the patient's state are repeated several times. Such situation always takes place in the controlling of the drug therapy efficacy. A specific feature of this diagnosis task is the dependence between patient's states at particular instants, which should be taken into account in sequential diagnosis algorithms. In this paper methods for performing sequential diagnosis using fuzzy sets and rough sets theory are developed and evaluated. For both soft methodologies several algorithms are proposed which differ in kind of input data and in details of classification procedures for particular instants of decision process. Proposed algorithms were practically applied to the computer-aided medical problem of recognition of patient's acid-base equilibrium states. Results of comparative experimental analysis of investigated algorithms in respect of classification accuracy are also presented and discussed. [ABSTRACT FROM AUTHOR]

The exponential increase in publication rate of new articles is limiting access of researchers to relevant literature. This has prompted the use of text mining tools to extract key biological information. Previous studies have reported extensive modification of existing generic text processors to process biological text. However, this requirement for modification had not been examined. In this study, we have constructed Muscorian, using MontyLingua, a generic text processor. It uses a two-layered generalization-specialization paradigm previously proposed where text was generically processed to a suitable intermediate format before domain-specific data extraction techniques are applied at the specialization layer. Evaluation using a corpus and experts indicated 86-90% precision and approximately 30% recall in extracting protein-protein interactions, which was comparable to previous studies using either specialized biological text processing tools or modified existing tools. Our study had also demonstrated the flexibility of the two-layered generalization-specialization paradigm by using the same generalization layer for two specialized information extraction tasks. [ABSTRACT FROM AUTHOR]

We develop a technique to validate large-scale gene regulatory networks (GRN) by comparing with corresponding protein-protein interaction (PPI) networks. The GRN are obtained with Bayesian networks while PPI networks are obtained from database of known PPI interactions. We look for exact matches and then reduced networks by skipping one or more genes in GRN. We demonstrate our technique on expression profiles of differentially expressed genes in the S. cerevisiae cell cycle. We validate GRNs against a merged database of 53235 genes. The precisions of GRN obtained over all genes were from 0.82 to 0.95 in all the phases. In particular we realized that one-skip and two-skip model significantly improved accuracy of the GRN of different phases of cell cycle. [ABSTRACT FROM AUTHOR]

Microarray gene expression data is used to model differential activity in Gene Regulatory Networks (GRN) to elucidate complex cellular processes, though network modeling is susceptible to errors due to both noisy nature of gene expression data and platform bias. This intuitively provided the motivation for the development of an innovative technique, which effectively integrates GRN using cross-platform data to minimize the two aforementioned effects. This paper presents a GRN integration (GeNi) framework that fuses cross-platform GRN to remove platform and experimental bias using the Dempster Shafer Theory of Evidence. The proposed model estimates gene co-regulation strength by using mutual information and removes spurious co-regulations through data processing inequality. The method automatically adapts to the data distribution using Belief theory, which does not require a preset threshold to accept co-regulated links. GeNi is applied to identify common cancer-related regulatory links in ten different datasets generated by different microarray platforms including cDNA and Affymetrix arrays. Experimental results demonstrate that GeNi can be effectively applied for GRN reconstruction and cross-platform gene network fusion for any gene expression data. [ABSTRACT FROM AUTHOR]

The self-organizing map (SOM) is useful within bioinformatics research because of its clustering and visualization capabilities. The SOM is a vector quantization method that reduces the dimensionality of original measurement and visualizes individual tumor sample in a SOM component plane. The data is taken from cDNA microarray experiments on Diffuse Large B-Cell Lymphoma (DLBCL) data set of Alizadeh. The objective is to get the SOM to discover biologically meaningful clusters of genes that are active in this particular form of cancer. Despite their powers of visualization, SOMs cannot provide a full explanation of their structure and composition without further detailed analysis. The only method to have gone someway towards filling this gap is the unified distance matrix or U-matrix technique. This method will be used to provide a better understanding of the nature of discovered gene clusters. We enhance the work of previous researchers by integrating the clustering results with the Gene Ontology for deeper analysis of biological meaning, identification of diversity in gene expression of the DLBCL tumors and reflecting the variations in tumor growth rate. [ABSTRACT FROM AUTHOR]

We have proposed an ant-based clustering algorithm for document clustering based on the travelling salesperson scenario. In this paper, we presented an approach called Ant-MST for gene expression data clustering based on both ant-based clustering and minimum spanning trees (MST). The ant-based clustering algorithm is firstly used to construct a fully connected network of nodes. Each node represents one gene, and every edge is associated with a certain level of pheromone intensity describing the co-expression level between two genes. Then MST is used to break the linkages in order to generate clusters. Comparing to other MST-based clustering approaches, our proposed method uses pheromone intensity to measure the similarity between two genes instead of using Euclidean distance or correlation distance. Pheromone intensities associated with every edge in a fully-connected network records the collective memory of the ants. Self-organizing behavior could be easily discovered through pheromone intensities. Experimental results on three gene expression datasets show that our approach in general outperforms the classical clustering methods such as K-means and agglomerate hierarchical clustering. [ABSTRACT FROM AUTHOR]

The gene-label correlation provides an effective measure of the relevancy of a gene. However, this measure evaluates genes on an individual basis, and the gene sets thus obtained may exhibit severe redundancy. In this study, we propose a new correlation heuristic for set-based gene selection, with the goal of alleviating the redundancy problem. The new correlation heuristic consists of two components that account for gene relevancy and redundancy respectively. The relevancy of a gene is evaluated in terms of its correlation with class label on an individual basis, while the redundancy of a gene with respect to a given gene subset is measured by its correlation with a new dimension built upon the gene subset. The new correlation heuristic retains the simplicity of individual gene evaluation and the redundancy handling capacity of set-based gene evaluation. Two different ways of using the relevancy and redundancy measures are presented in this study. One way is the maximization of the ratio of relevancy measure to redundancy measure, and another way is the maximization of the relevancy measure subtracting redundancy measure. Experimental studies on six gene expression problems show that both criteria produce excellent results. [ABSTRACT FROM AUTHOR]

This paper concerns the discovery of Order Preserving Clusters (OP-Clusters) in gene expression data, in each of which a subset of genes induce a similar linear ordering along a subset of conditions. After converting each gene vector into an ordered label sequence. The problem is transferred into finding frequent orders appearing in the sequence set. We propose an algorithm of finding the frequent orders by iteratively Combining the most Frequent Prefixes and Suffixes (CFPS) in a statistical way. We also define the significance of an OP-Cluster. Our method has good scale-up property with dimension of the dataset and size of the cluster. Experimental study on both synthetic datasets and real gene expression dataset shows our approach is very effective and efficient. [ABSTRACT FROM AUTHOR]

Many advanced machine learning and statistical methods have recently been employed in classification of gene expression measurements. Although many of these methods can achieve high accuracy, they generally lack comprehensibility of the classification process. In this paper a new method for interpretation of small ensembles of classifiers is used on gene expression data from real-world dataset. It was shown that interactive interpretation systems that were developed for classical machine learning problems also give a great range of possibilities for the scientists in the bioinformatics field. Therefore we chose a gene expression dataset discriminating three types of Leukemia as a testbed for the proposed Visual Interpretation of Small Ensembles (VISE) tool. Our results show that using the accuracy of ensembles and adding comprehensibility gains not only accurate but also results that can possibly represent new knowledge on specific gene functions. [ABSTRACT FROM AUTHOR]

DNA Microarray technology allow us to identify cancerous tissues considering the gene expression levels across a collection of related samples. Several classifiers such as Support Vector Machines (SVM), k Nearest Neighbors (k-NN) or Diagonal Linear Discriminant Analysis (DLDA) have been applied to this problem. However, they are usually based on Euclidean distances that fail to reflect accurately the sample proximities. Several classifiers have been extended to work with non-Euclidean dissimilarities although none outperforms the others because they misclassify a different set of patterns. In this paper, we combine different kind of dissimilarity based classifiers to reduce the misclassification errors. The diversity among classifiers is induced considering a set of complementary dissimilarities for three different type of models. The experimental results suggest that the algorithm proposed helps to improve classifiers based on a single dissimilarity and a widely used combination strategy such as Bagging. [ABSTRACT FROM AUTHOR]

We have developed an online database APMA (Affymetrix Probe Mapping and Annotation) for interactive presentation, search and visualization of Affymetrix target sequences mapping and annotation . APMA contains revised genome localization of the Affymetrix U133 GeneChip initial (target) probe sequences. We designed APMA to use it as a filter before data analysis and data mining so that noise expression signals, false correlations and false gene expression patterns can be reduced. Discrepancies found in probeset annotation and target sequence mapping account for up to 30% of probesets, including about 25% of Affymetrix probesets derived from target sequences overlapped interspersed repeats and 1.8% of original target sequences with erroneous orientation of the sequences. 86% of U133 target sequences passed our quality-control filtering. [ABSTRACT FROM AUTHOR]

The analysis on the amino acid sequences of transmembrane beta barrel proteins (TMBs) provides deep insights about their structure and function. We found that the occurrence of Ser, Asn and Gln is significantly higher in TMBs than globular proteins, which might be due to their importance in the formation of β-barrel structures in the membrane, stability of binding pockets and the function of TMBs. Utilizing this information, we have devised statistical methods and machine learning techniques to discriminate TMBs from other folding types of globular and membrane proteins and we obtained the maximum accuracy of 96%. Further, we have devised protocols for identifying the membrane spanning β-strand segments and detecting TMBs in genomic sequences. [ABSTRACT FROM AUTHOR]

Hydrophobicity has long been considered as one of the primary driving forces in the folding of proteins. We discuss here the evolutionary average of the hydrophobicity profile in an aligned family of proteins and found a patchy mean hydrophobicity profile. This is in contrast to Bastolla et al (2005b) results for the large superfamily of globular proteins. The idea is to use singular value decomposition and cavity filtering in order to remove the eigensequences burried in the evolutionary noise [ABSTRACT FROM AUTHOR]

Many structural and functional properties of proteins can be described as a one-dimensional one-to-one mapping between residues of protein sequence and target structure or function. These residue level properties (RLPs) have been frequently predicted using neural networks and other machine learning algorithms. Here we present an algorithm to dynamically exclude from the neural network training, examples which are most difficult to separate. This algorithm automatically filters out statistical outliers causing noise and makes training faster without losing network ability to generalize. Different methods of sampling data for neural network training have been tried and their impact on learning has been analyzed. [ABSTRACT FROM AUTHOR]

The right-handed single-stranded ß-helix proteins are characterized as virulence factors, allergens, toxins that are threat to human health. Identification of these proteins from amino acid sequence is of great importance as these proteins are potential targets for anti-bacterial and fungal agents. In this paper, support vector machine (SVM) has been used to predict the presence of ß-helix fold in protein sequences using dipeptide composition. An input vector of 400 dimensions for dipeptide compositions is used to search for the presence of putative rungs or coils, the conserved secondary structure, found in ß-helix proteins. An average accuracy of 89.2% with Matthew's correlation coefficient of 0.75 is obtained in a 5-fold cross-validation technique. In addition, a PSSM was also used to score the query sequence of proteins identified as ß-helices by SVM. The method recognizes right-handed ß-helices with 100% sensitivity and 99.6% specificity on test set of known protein structures. [ABSTRACT FROM AUTHOR]

Protein-protein interactions play a very important role in many biological processes, for example, information transfer along signaling pathways, and enzyme catalysis. Recently, scientists tried to predict the protein-protein interaction interface from sequences. Since the number of protein 3D structure still increase slowly comparing to the number of protein sequences, it may be a good idea to predict the protein-protein interface from sequences directly. In this paper, the compositions and conserved functions of the amino acids in the protein interface are studied, and the information of secondary structures is added. In addition, we used radio basis function network to predict the protein interface with adding some useful biochemical features. [ABSTRACT FROM AUTHOR]

We have investigated the relative performance of amino acid occurrence and other features, such as predicted secondary structure, hydrophobicity, normalized van der Waals volume, polarity, polarizability, and real/predicted contact information of residues, for recognizing protein folds. We observed that the improvement over other features is only marginal compared with amino acid occurrence. This is because amino acid occurrence, indirectly, can consider varieties of physical properties which are useful to discriminate protein folds. If we consider only proteins which are well aligned structurally with each other, the accuracy of discrimination is drastically improved. In order to discriminate protein folds more accurately, we need to consider anything other than structure alignment. [ABSTRACT FROM AUTHOR]

Prediction of protein structural classes for low homology proteins is a challenging research task in bioinformatics. A dual-layer fuzzy support vector machine (FSVM) network approach is proposed to predict protein structural classes. A protein sample can be represented by nine representation feature vectors: pair couple amino acid (210-D) and eight pseudo amino acid composition vectoers (PseAAC). Eight physicochemical properties of amino acids extracted from AAIndex databank are used to calculate low frequencies of power spectrum density of sequence-order correlation in protein sequence. In the first layer of FSVM network, nine FSVM classifiers are established, which are trained by different protein feature vectors, respectively. The outputs of the first layer are reclassified by FSVM classifier in 2nd layer of the network. The performance of proposed method is validated by low homology (average 25%) dataset covering 1673 proteins. The promising results indicate that the new method may become a useful tool for predicting not only the structural classification of proteins but also their other attributes. [ABSTRACT FROM AUTHOR]

The schemata theorem, on which the working of Genetic Algorithm (GA) is based in its current form, has a fallacious selection procedure and incomplete crossover operation. In this paper, generalization of the schemata theorem has been provided by correcting and removing these limitations. The analysis shows that similarity growth within GA population is inherent due to its stochastic nature. While the stochastic property helps in GA's convergence. The similarity growth is responsible for stalling and becomes more prevalent for hard optimization problem like protein structure prediction (PSP). While it is very essential that GA should explore the vast and complicated search landscape, in reality, it is often stuck in local minima. This paper shows that, removal of members of population having certain percentage of similarity would keep GA perform better, balancing and maintaining convergence property intact as well as avoids stalling. [ABSTRACT FROM AUTHOR]

Theoretical and computational methods for the prediction of protein subcellular localization have been proposed and are developing continuously. Many representations of protein sequence are proposed but a new problem arises: how to organize them together to improve prediction. It is an available solution to serialize multiple representations to single bigger one, but is still hard to avoid calculation error derived from greatly different feature values and causes huge computational burden natively because of high dimensional feature vector. We present a novel method based on decision templates(DT) for such problems in this paper. First, a protein sequence is represented as three new types of feature vectors. Then, the feature vectors are further taken as the inputs of individual SVM classifiers respectively. Finally, the outputs of these classifiers are aggregated by decision templates. The results demonstrate that DT is superior to other methods of subcellular localization prediction. [ABSTRACT FROM AUTHOR]

This work performs an analysis on two, quite different, techniques for Quantitative Trait Loci (QTL) Analysis. Interval Mapping (IM) as described by Karl Broman is compared to a Hierarchical Bayesian Model (HBM) technique that reduces the problem of QTL analysis down to one of model selection. Simulations were generated for the flowering plant of the Arabidopsis thaliana for evaluation of the techniques. It is shown that the HBM technique was much more successful at determining the appropriate loci/markers and corresponding chromosomes than the IM technique given a single loci. It was further elucidated through simulation runs that the HBM was robust against two loci/markers, whereas IM completely failed. The contribution of this work is in the comparison and analysis of the IM method to that of the HBM; hence, demonstrating through simulations that the HBM technique is superior to that of the IM for the Arabidopsis simulated data. [ABSTRACT FROM AUTHOR]

Chicken genome AT and GC skews for individual chromosomes were visualized simultaneously using a novel method of 2-dimensional color-coded pixel matrix. The visualizations were compared to those of human, mouse and possum genomes. A strikingly strong correlation of AT skew and GC from small to large scale in chicken genome was found, compared to the other vertebrates. Some local skew correlations were also found for the other vertebrates, but only in small genomic scale. Quantitative measures of correlation were developed, and confirmed the special characteristic of chicken chromosomes. Possible explanations for uniqueness of birds in this respect are discussed. The phylogenetic distribution and evolutionary pressures responsible for this previously unreported skew correlation warrant further study. [ABSTRACT FROM AUTHOR]

The human adenosine A2a receptor (A2aR) belongs to the family of G-protein coupled receptors (GPCRs), characterized by seven transmembrane (TM) helices. TMs are involved in various cellular processes including dimerization-mediated recognition of ligand. TM5 has been suggested to self associate and may be involved in the dimerization of A2aR. However the role of dimerization and the motifs involved in dimerization of TM 5 have not been revealed. To study the folding and assembly of A2aR, the cDNA of the adenosine A2aR from rat brain was isolated and sequenced (DQ098650). The computational analysis (gi70727927gbAAZ07991.1) showed that the protein of 42 amino acid residues aligned in TM 5 domain region of AA2AR_RAT (P30543). PROSITE search illustrated that the motif PMNYM was conserved in A2aR and the motif PMSYM was present in A2bR respectively. The minimal dimerization motif in the TM 5 domain of the rat A2a receptor sequence DQ098650 has found to be the motif PXXXM/Y. [ABSTRACT FROM AUTHOR]

Human genetics is undergoing a data explosion. Methods are available to measure DNA sequence variation throughout the human genome. Given current knowledge it seems likely that common human diseases are best predicted by interactions between biological components, which can be examined as interacting DNA sequence variations. The challenge is thus to examine these high-dimensional datasets to identify combinations of variations likely to predict common diseases. The goal of this paper was to develop and evaluate a genetic programming (GP) mutator suited to this task by exploiting expert knowledge in the form of Tuned ReliefF (TuRF) scores during mutation. We show that using expert knowledge guided mutation performs similarly to expert knowledge guided selection. This study demonstrates that in the context of an expert knowledge aware GP, mutation may be an appropriate component of the GP used to search for interacting predictors in this domain. [ABSTRACT FROM AUTHOR]

Steroid hormone receptors compose a subgroup of regulatory proteins which tend to recognize partially symmetric response elements on DNA. Identification of the members of a gene regulatory machine conducted by steroid hormones could provide better understanding of nature and development of diseases. We present an approach based on a succession of neural networks, which can be used for highly specific detection of binding signals. It exploits the capability of a feed-forward neural network to model datasets with high confidence, while a recurrent network grants putative response elements with biologically meaningful structures. We have used a novel method to train such a two-phase artificial neural network with a set of experimentally validated response elements for steroid hormone receptors. We have demonstrated that sequence-based prediction followed by structure-based classification of putative binding sites allows to eliminate large amount of false positives. An implementation of the neural network with Field-Programmable Gate Array is also briefly described. [ABSTRACT FROM AUTHOR]

Systolisation of the pairwise distance computation algorithm and mapping into field programmable gate arrays (FPGA) have proven to give superior performance at a lower cost, compared to the same algorithm running on a cluster of workstations. The primary design methodology for this approach is based on the hardware description languages such as VHDL and Verilog HDL. An alternative design methodology, however, is the use of a high level language such as C to describe the algorithms and generate equivalent hardware descriptions for implementation in FPGA so as to reduce time to market. This paper describes the design and implementation of the ClustalW first stage multiple sequence alignment based on the Smith-Waterman algorithm on a low cost FPGA development platform using a C language development tool suite. Performance evaluation results show that comparable performance could be achieved to that of Pentium 4 systems and other HDL-based solutions using even the smallest commercially available FPGA device with this design methodology. [ABSTRACT FROM AUTHOR]

With the extraordinarily increase in genomic sequence data, there is a need to develop an effective and accurate method to deduce the biological functions of novel sequences with high accuracy. As the use of experiments to validate the function of biological sequence is too expensive and hardly to be applied to large-scale data, the use of computer for prediction of gene function has become an economical and effective substitute. This paper proposes a new design of BLAST-based GO term annotator which incorporates data mining techniques and utilizes rough set theory. Moreover, this method is an evolution against the traditional methods which only base on BLAST or characters of GO Terms. Finally, experimental results prove the validity of the proposed rough set-based method. [ABSTRACT FROM AUTHOR]

This article presents a continuation of our research aiming at improving the effectiveness of signal decomposition algorithms by providing them with "classification-awareness." We investigate hybridization of multi-objective evolutionary algorithms (MOEA) and rough sets (RS) to perform the task of decomposition in the light of the underlying classification problem itself. In this part of the study, we also investigate the idea of utilizing the Independent Component Analysis (ICA) to initialize the population in the MOEA. [ABSTRACT FROM AUTHOR]

Cellular and molecular imaging promises powerful tools for the visualization and elucidation of important disease-causing biological processes. Recent research aims to simultaneously assess the spatial-spectral/temporal distributions of multiple biomarkers, where the signals often represent a composite of more than one distinct source independent of spatial resolution. We report here a blind source separation method for quantitative dissection of mixed yet correlated biomarker patterns. The computational solution is based on a latent variable model, whose parameters are estimated using the non-negative least-correlated component analysis (nLCA) proposed in this paper. We demonstrate the efficacy of the nLCA with real bio-imaging data. With accurate and robust performance, it has powerful features which are of considerable widespread applicability. [ABSTRACT FROM AUTHOR]

Biological research frequently requires specialist databases to support in-depth analysis about specific subjects. With the rapid growth of biological sequences in public domain data sources, it is difficult to keep these databases current with the sources. Simple queries formulated to retrieve relevant sequences typically return a large number of false matches and thus demanding manual filtration. In this paper, we propose a novel methodology that can support automatic incremental updating of specialist databases. Complex queries for incremental updating of relevant sequences are learned using Association Rule Mining (ARM), resulting in a significant reduction in false positive matches. This is the first time ARM is used in formulating descriptive queries for the purpose of incremental maintenance of specialised biological databases. We have implemented and tested our methodology on two real-world databases. Our experiments conclusively show that the methodology guarantees an F-score of up to 80% in detecting new sequences for these two databases. [ABSTRACT FROM AUTHOR]

Many entries in major biological databases have incomplete functional annotation and thus, frequently, it is difficult to identify entries for a specific functional category. We combined information of protein functional domains and gene ontology descriptions for highly accurate identification of transcription factor (TF) entries in Swiss-Prot and Entrez Gene databases. Our method utilizes support vector machines and it efficiently separates TF entries from non-TF entries. The 10-fold cross validation of predictions produced on average a positive predictive value of 97.5% and sensitivity of 93.4%. Using this method we have scanned the whole Swiss-Prot and Entrez Gene databases and extracted 13826 unique TF entries. Based on a separate manual test of 500 randomly chosen extracted TF entries, we found that the non-TF (erroneous) entries were present in 2% of the cases. [ABSTRACT FROM AUTHOR]

Two promising approaches for handling large-scale biodata are presented and illustrated in several new contexts: molecular structure bitmap image processing for chemoinformatics, and fractal visualization methods for genome analyses. It is suggested that two-dimensional structure databases of bioactive molecules (e.g. proteins, drugs, folded RNAs), transformed to bitmap image databases, can be analysed by a variety of image processing methods, with an example of human microRNA folded 2D structures processed by Gabor filter. Another compact and efficient visualization method is comparison of huge amounts of genomic and proteomic data through fractal representation, with an example of analyzing oligomer frequencies in a bacterial phytoplasma genome. Bitmap visualization of bioinformatics data seems promising for complex parallel pattern discovery and large-scale genome comparisons, as powerful modern image processing methods can be applied to the 2D images. [ABSTRACT FROM AUTHOR]

Our goal is to develop a state-of-the-art secondary structure predictor with an intuitive and biophysically-motivated energy model through the use of Hidden Markov Support Vector Machines (HM- SVMs), a recent innovation in the field of machine learning. We focus on the prediction of alpha helices and show that by using HM-SVMs, a simple 7-state HMM with 302 parameters can achieve a Qα value of 77.6% and a SOVα value of 73.4%. As detailed in an accompanying technical report[11], these performance numbers are among the best for techniques that do not rely on external databases (such as multiple sequence alignments). [ABSTRACT FROM AUTHOR]

When the same set of genes appear in different orders on the chromosomes, they form a permutation pattern. Permutation patterns have been used to identify potential haplogroups in mammalian data [8]. They also have been successfully used to detect phylogenetic relationships between computer viruses [9]. In this paper we explore the use of these patterns as a content similarity measure and use this in inferring phylogenies from genome rearrangement data in polynomial time. The method uses a function of the cardinality of the set of common maximal permutation patterns as a proxy for evolutionary "proximity" between genomes. We introduce Pi-logen, a phylogeny tool based on this method. We summarize results of feasibility study for this scheme on synthetic data by (1) content verification and (2) ancestor prediction. We also successfully infer phylogenies on series of synthetic data and on chloroplast gene order of Campanulaceae data. [ABSTRACT FROM AUTHOR]

Epitopes are defined as the molecular structures interacting with specific receptors of the immune system such as antibodies, MHC, and T cell receptor molecules. The Immune Epitope Database and Analysis Resource (IEDB, http://www.immuneepitope.org) is a database specifically devoted to immune epitope data. The database is populated with intrinsic and context-dependent epitope data curated from the scientific literature by immunologists, biochemists, and microbiologists. An analysis resource is linked to the database which hosts various bioinformatics tools to analyze epitope data as well as to predict de novo epitopes. The availability of the IEDB will facilitate the exploration of immunity to infectious diseases, allergies, autoimmune diseases, and cancer. The utility of the IEDB was recently demonstrated through a comprehensive analysis of all current information regarding antibody and T cell epitopes derived from influenza A and determining possible cross-reactivity among H5N1 avian flu and human flu viruses. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the topological properties of synthetic RNAs (i.e., functional RNAs synthesized by in vitro selection technique), by applying the spectral graph partitioning technique. Our analysis shows that the majority of synthetic RNAs possess between two to six vertices and their second eigenvalues lie between one and two. In contrast, natural RNA structures mostly have nine or ten vertices and are less compact with the second eigenvalue below unity. Our statistical analysis (at 95 percentile) also reveals three criteria important for designing novel functional RNAs. Firstly, RNA sequences screened from a large random library, with length of 80 nucleotides and 32.31% paired bases, are very likely to fold into functional RNAs. Secondly, their predicted structures should possess two to six vertices inclusively. Thirdly, to minimize the number of false positives, a combination of filtering parameters should be included, the percentage G/C content of 65.95% and the normalized minimum free energy of -0.021 kcal/mol per nucleotide. [ABSTRACT FROM AUTHOR]

Feature selection is crucial to tumor classification due to the high dimensionality of microarray datasets. With the aid of the degree of differential prioritization (DDP) between relevance and antiredundancy, our proposed DDP-based feature selection technique is capable of achieving better accuracies than those reported in previous studies, while using fewer genes in the predictor set. Additionally, we discovered a strong correlation between the DDP parameter in our feature selection technique and the number of classes in the dataset. This leads us to question if the measure of relevance in our feature selection technique becomes less efficient at capturing the class-specific relevance for each individual class of the dataset as the number of classes increases. In this study, we analyze the class-specific relevance of the predictor sets found using our feature selection technique. The analysis ultimately lays down the theoretical foundation for a beneficial improvement to our feature selection technique. [ABSTRACT FROM AUTHOR]

We propose a novel mixture probability model for the probability distribution function (PDF) of microarray signals, which comprises a noise and a signal component. The noise term, due to non-specific mRNA hybridization, is given by a lognormal distribution; and the true signal, from specific mRNA hybridization, is described by the generalized Pareto-gamma (GPG) function. The model, applied to expression data of 251 human breast cancer tumors on the Affymetrix microarray platform, yields accurate fits for all tumor samples. We observe that (i) high aggressive cancers have, in general, broader right tails in the GPG than low aggressive cancers; (ii) the exponent parameter value of the GPG distribution is not constant and correlates strongly with ~4000 expressed genes and several "gold standard" clinical risk factors. These results can not be obtained from so-called "scale-free network" models. We conclude that an accurate parameterization of scale-dependent GPG function could provide robust prognostic benefits for cancer patients. [ABSTRACT FROM AUTHOR]

As more and more genomes have been discovered in recent years, it is an urgent need to develop a reliable method to predict protein subcellular localization for further function exploration. However many well-known prediction methods based on amino acid composition, have no ability to utilize the information of sequence-order. Here we propose a novel method, named moment descriptor (MD), which can obtain sequence order information in protein sequence without the need of the information of physicochemical properties of amino acids. The presented method first constructs three types of moment descriptors, and then applies multi-class SVM to the Chou's dataset. Through resubstitution, jackknife and independent tests, it is shown that the MD is better than other methods based on various types of extensions of amino acid compositions. Moreover, three multi-class SVMs show similar performance except for the training time. [ABSTRACT FROM AUTHOR]