Catálogo de publicaciones - libros

Protein-protein interaction networks, particularly that of the yeast , have recently been studied extensively. These networks seem to satisfy the small world property and their (1-hop) degree distribution seems to form a power law. More recently, a number of duplication based random graph models have been proposed with the aim of emulating the evolution of protein-protein interaction networks and satisfying these two graph theoretical properties. In this paper, we show that the proposed model of Pastor-Satorras et al. does not generate the power law degree distribution with exponential cutoff as claimed and the more restrictive model by Chung et al. cannot be interpreted unconditionally. It is possible to slightly modify these models to ensure that they generate a power law degree distribution. However, even after this modification, the more general k-hop degree distribution achieved by these models, for  > 1, are very different from that of the yeast proteome network. We address this problem by introducing a new network growth model that takes into account the sequence similarity between pairs of proteins (as a binary relationship) as well as their interactions. The new model captures not only the k-hop degree distribution of the yeast protein interaction network for all  > 0, but it also captures the 1-hop degree distribution of the sequence similarity network, which again seems to form a power law.

We conduct the transcription factor (TF) analysis by detecting transcription factor pairs and incorporating binding positions for genes with altered expressions in time-series cDNA microarray data. Prediction of TF pairs that mostly likely contribute to the regulated transcription of differentially expressed genes are done through the computation of their expression coherence (EC). The Expectation Maximization (EM) clustering is performed additionally in order to detect patterns in specific TF binding positions. We evaluate the EC of expression profiles of genes within each cluster to discover binding trends that may play a significant role in regulation of target genes. Our method has successfully identified TF pairs that have a greater potential for regulating their target genes at specified locations rather than at arbitrary locations.

A series of genome-scale algorithms and high-performance implementations is described and shown to be useful in the genetic analysis of gene transcription. With them it is possible to address common questions such as: “are the sets of genes co-expressed under one type of conditions the same as those sets co-expressed under another?” A new noise-adaptive graph algorithm, dubbed “paraclique,” is introduced and analyzed for use in biological hypotheses testing. A notion of vertex coverage is also devised, based on vertex-disjoint paths within correlation graphs, and used to determine the identity, proportion and number of transcripts connected to individual phenotypes and quantitative trait loci (QTL) regulatory models. A major goal is to identify which, among a set of candidate genes, are the most likely regulators of trait variation. These methods are applied in an effort to identify multiple-QTL regulatory models for large groups of genetically co-expressed genes, and to extrapolate the consequences of this genetic variation on phenotypes observed across levels of biological scale through the evaluation of vertex coverage. This approach is furthermore applied to definitions of homology-based gene sets, and the incorporation of categorical data such as known gene pathways. In all these tasks discrete mathematics and combinatorial algorithms form organizing principles upon which methods and implementations are based.

The stability and fragility of four species’ protein-protein interaction networks (PINs) are studied by investigating their robustness, i.e. their topological parameters retain a similar system behavior with respect to four different types of perturbations. Four types of perturbations are considered; that is (i) network nodes are randomly removed (failure), (ii) the most connected node is successively removed (attack), (iii) interaction edges are rewired randomly, and (iv) edges are randomly deleted. At most 50% of network nodes or edges are deleted or rewired. It is demonstrated that PINs are quite robust with respect to failure, attack, random rewiring and edge deletion, that is the average diameters for perturbed networks differ from the unperturbed cases have a difference less than 13%, which is relative small in comparison with WWW and the Internet results. These results suggest that PINs’ network topologies are robust with respect to perturbations.

The advent of high-throughput techniques is transforming biology into a data rich field. A variety of genomics data is now available, each providing a different perspective of gene regulation. Even though each type of data requires specific computational methods, methods that combine complimentary datasets are necessary to obtain additional information that is not available by analyzing the either of the dataset alone. In this paper, we propose a Bayesian approach to integrate gene expression data with genome-wide protein-DNA interaction data. The proposed method combines these datasets in order to probabilistic predict transcription factors for genes. We evaluate the proposed method using Saccharomyces Cell Cycle data. Results are compared with that of previous method.

Peptide recognition modules (PRMs) are specialised compact protein domains that mediate many important protein-protein interactions. They are responsible for the assembly of critical macromolecular complexes and biochemical pathways [Pawson and Scott, 1997], and they have been implicated in carcinogenesis and various other human diseases [Sudol and Hunter, 2000]. PRMs recognise and bind to peptide ligands that contain a specific structural motif. This paper introduces a novel discriminative model which models these PRMs and allows prediction of their behaviour, which we compare with a recently proposed generative model. We find that on a yeast two-hybrid dataset, the generative model performs better when background sequences are included, while our discriminative model performs better when the evaluation is focused on discriminating between the SH3 domains. Our model is also evaluated on a phage display dataset, where it consistently out-performed the generative model.

In order to guarantee that the optimal motif is found, traditional pattern-driven approaches perform an exhaustive search over all candidate motifs of length . We develop an improved pattern-driven algorithm that takes (4) time, where is the number of sequences in the sample and is the motif length, which is independent of the length of each sequence for large enough and saving a factor of in time complexity over the original pattern-driven approach. We further extend this strategy to allow arbitrary don’t care positions within a motif without much decrease in solvable values of . Testing this algorithm on a large set of yeast samples constructed from co-expressed gene clusters reveals that most biological motifs have many invariant or almost invariant positions and these positions can be used to define the motif while ignoring the other positions. This motivates the following two-stage strategy that extends the solvable values of substantially for the pattern-driven approach: first use an (2) algorithm to exhaustively search over all candidate motifs allowing arbitrary don’t care positions but disallowing mismatches, then refine these motifs by allowing a limited amount of flexibility to model the almost invariant positions. We demonstrate that this seemingly restrictive motif definition is sufficiently powerful by showing that the performance of this algorithm is comparable to the best existing motif finding algorithms on a large benchmark set of samples. A software program implementing these approaches (MotifEnumerator) is available at http://faculty.cs.tamu.edu/shsze/motifenumerator.

Understanding the attributes that confer promoter activity is essential for gene regulation, gene prediction and sequence annotation. It is a challenging problem to detect promoter regions, either or by experiments. In this report, we show that the stress induced DNA duplex destabilization sites (SIDD) in prokaryotic genomes under negative superhelical stresses, as occurs , are closely associated with specific promoter regions. When compared with DNA curvature, deformability, thermostability or sequence motif scores within the -10 region, SIDD is the most informative DNA property of promoter regions in the K12 genome. Our method using SIDD as a sole predictor performs better than other promoter prediction programs in detecting promoter sequences in or . We show that, by combining SIDD properties with -10 motif scores in a linear discrimination function, one can achieve better than 80% accuracy in predicting promoter sequences.

Transcriptional regulation is accomplished by several transcription factor proteins that bind to specific DNA elements in the relative vicinity of the gene, and interact with each other and with Polymerase enzyme. Thus the determination of transcription factor-DNA binding is an important step toward understanding transcriptional regulation. An effective way to experimentally determine the genomic regions bound by a transcription factor is by a ChIP-on-chip assay. Then, given the putative genomic regions, computational motif finding algorithms are applied to estimate the DNA binding motif or positional weight matrix for the TF. The expectation is that the presence or absence of the estimated motif in a promoter should be a good indicator of the binding of the TF to that promoter. This association between the presence of the transcription factor motif and its binding is however weak in a majority of cases where the whole genome ChIP experiments have been performed. One possible reason for this is that the DNA binding of a particular transcription factor depends not only on its own motif, but also on synergistic or antagonistic action of neighboring motifs for other transcription factors. We believe that modeling this interaction-dependent binding with linear regression can better explain the observed binding data. We assess this hypothesis based on the whole genome ChIP-on-chip data for Yeast. The derived interactions are largely consistent with previous results that combine ChIP-on-chip data with expression data. We additionally apply our method to determine interacting partners for CREB and validate our findings based on published experimental results.

MicroRNAs are short, noncoding RNAs that play important roles in post-transcriptional regulation. Although many functions of microRNAs in plants and animals have been revealed in recent years, the transcriptional mechanism of microRNA genes is not well understood. To elucidate the transcriptional regulation of microRNA genes, we study and characterize, in a genome scale, the promoters of intergenic microRNA genes in , and . Specifically, we show that the known microRNA genes in these species have the same type of promoters as the protein-coding genes. To further characterize the promoters of miRNA genes, we develop a miRNA core promoter prediction method, called (CoVote). We applied this new method to identify putative core promoters of most known microRNA genes in the three model species of choice.