Catálogo de publicaciones - libros

Differential equations have been established to model the dynamic behavior of gene regulatory networks in the last few years. They provide a detailed insight into regulatory processes at a molecular level. However, in a top down approach aiming at the inference of the underlying regulatory network from gene expression data, the corresponding optimization problem is usually severely underdetermined, since the number of unknowns far exceeds the number of timepoints available. Thus one has to restrict the search space in a biologically meaningful way.

We use differential equations to model gene regulatory networks and introduce a Bayesian regularized inference method that is particularly suited to deal with sparse and noisy datasets. Network inference is carried out by embedding our model into a probabilistic framework and maximizing the posterior probability. A specifically designed hierarchical prior distribution over interaction strenghts favours sparse networks, enabling the method to efficiently deal with small datasets.

Results on a simulated dataset show that our method correctly learns network structure and model parameters even for short time series. Furthermore, we are able to learn main regulatory interactions in the yeast cell cycle.

The paper presents MRNet, an original method for inferring genetic networks from microarray data. This method is based on Maximum Relevance – Minimum Redundancy (MRMR), an effective information-theoretic technique for feature selection.

MRNet is compared experimentally to Relevance Networks (RelNet) and ARACNE, two state-of-the-art information-theoretic network inference methods, on several artificial microarray datasets. The results show that MRNet is competitive with the reference information-theoretic methods on all datasets. In particular, when the assessment criterion attributes a higher weight to precision than to recall, MRNet outperforms the state-of-the-art methods.

A gene regulatory network (GRN) extracted from microarray data has the potential to give us a concise and precise way of understanding how genes interact under the experimental conditions studied [1, 2]. Learning such networks, and unravelling the knowledge hidden within them is important for drug targets and to understand the basis of disease. In this paper, we analyse microarray gene expression data from to extract Bayesian belief networks (BBNs) which mirror the cell cycle GRN. This is achieved through the use of a novel structure learning algorithm of Taboo search and a novel knowledge extraction technique, target node (TN) analysis. We also show how quantitative and qualitative information captured within the BBN can be used to simulate the nature of interaction between genes. The GRN extracted was validated against literature and genomic databases, and found to be in excellent agreement.

Experimental time courses often reveal a nonlinear behaviour. Analysing these nonlinearities is even more challenging when the observed phenomenon is cyclic or oscillatory. This means, in general, that the data describe a circular trajectory which is caused by periodic gene regulation.

Nonlinear PCA (NLPCA) is used to approximate this trajectory by a curve referred to as nonlinear component. Which, in order to analyse cyclic phenomena, must be a closed curve hence a circular component. Here, a neural network with circular units is used to generate circular components.

This circular PCA is applied to gene expression data of a time course of the intraerythrocytic developmental cycle (IDC) of the malaria parasite . As a result, circular PCA provides a model which describes continuously the transcriptional variation throughout the IDC. Such a computational model can then be used to comprehensively analyse the molecular behaviour over time including the identification of relevant genes at any chosen time point.

Freezing tolerance in plants is obtained during a period of low non-freezing temperatures before the winter sets on, through a biological process known as cold acclimation. Cold is one of the major stress factors that limits the growth, productivity and distribution of plants, and understanding the mechanism of cold tolerance is therefore important for crop improvement. Expressed sequence tags (EST) analysis is a powerful, economical and time-efficient way of assembling information on the transcriptome. To date, several EST sets have been generated from cold-induced cDNA libraries from several different plant species. In this study we utilize the variation in the frequency of ESTs sampled from different cold-stressed plant libraries, in order to identify genes preferentially expressed in cold in comparison to a number of control sets. The species included in the comparative study are oat (), barley (), wheat (), rice () and . However, in order to get comparable gene expression estimates across multiple species and data sets, we choose to compare the expression of tentative ortholog groups (TOGs) instead of single genes, as in the normal procedure. We consider TOGs as preferentially expressed if they are detected as differentially expressed by a test statistic and up-regulated in comparison to all control sets, and/or uniquely expressed during cold stress, i.e., not present in any of the control sets. The result of this analysis revealed a diverse representation of genes in the different species. In addition, the derived TOGs mainly represent genes that are long-term highly or moderately expressed in response to cold and/or other stresses.

We analyse data from the Edinburgh Mouse Atlas Gene-Expression Database (EMAGE) which is a high quality data source for spatio-temporal gene expression patterns. Using a novel process whereby generated patterns are used to probe spatially-mapped gene expression domains, we are able to get unbiased results as opposed to using annotations based predefined anatomy regions. We describe two processes to form association rules based on spatial configurations, one that associates spatial regions, the other associates genes.

It has previously been demonstrated that gene expression data correlate with event-free and overall survival in several cancers. A number of methods exist that assign patients to different risk classes based on expression profiles of their tumor. However, predictions of actual survival times in years for the individual patient, together with confidence intervals on the predictions made, would provide a far more detailed view, and could aid the clinician considerably in evaluating different treatment options. Similarly, a method able to make such predictions could be analyzed to infer knowledge about the relevant disease genes, hinting at potential disease pathways and pointing to relevant targets for drug design. Here too, confidences on the relevance values for the individual genes would be useful to have.

Our algorithm to tackle these questions builds on a hierarchical Bayesian approach, combining a Cox regression model with a hierarchical prior distribution on the regression parameters for feature selection. This prior enables the method to efficiently deal with the low sample number, high dimensionality setting characteristic of microarray datasets. We then sample from the posterior distribution over a patients survival time, given gene expression measurements and training data. This enables us to make statements such as . A similar approach is used to compute relevance values with confidence intervals for the individual genes measured.

The method is evaluated on a simulated dataset, showing feasibility of the approach. We then apply the algorithm to a publicly available dataset on diffuse large B-cell lymphoma, a cancer of the lymphocytes, and demonstrate that it successfully predicts survival times and survival time distributions for the individual patient.

In single-nucleotide polymorphism (SNP) association studies interactions are often of main interest. Logic regression is a regression methodology that can identify complex Boolean interactions of binary variables. It has been applied successfully to SNP data but only identifies a single best model, while usually there is a number of models that are almost as good. Extensions of logic regression that consider several plausible models are Monte Carlo logic regression (MCLR) and a full Bayesian version of logic regression (FBLR) proposed in this paper. FBLR allows the incorporation of biological knowledge such as known pathways. We compare the performance in identifying SNP interactions associated with the case-control status of the three logic regression based methods and stepwise logistic regression in a simulation study and in a study of breast cancer.

Investigation of 3D chromatin structure in interphase cell nuclei is important for the understanding of genome function. For a reconstruction of the 3D architecture of the human genome, systematic fluorescent in situ hybridization in combination with 3D confocal laser scanning microscopy is applied. The position of two or three genomic loci plus the overall nuclear shape were simultaneously recorded, resulting in statistical series of pair and triple loci combinations probed along the human chromosome 1 q-arm. For interpretation of statistical distributions of geometrical features (e.g. distances, angles, etc.) resulting from finite point sampling experiments, a Monte-Carlo-based approach to numerical computation of geometrical probability density functions (PDFs) for arbitrarily-shaped confined spatial domains is developed. Simulated PDFs are used as bench marks for evaluation of experimental PDFs and quantitative analysis of dimension and shape of probed 3D chromatin regions. Preliminary results of our numerical simulations show that the proposed numerical model is capable to reproduce experimental observations, and support the assumption of confined random folding of 3D chromatin fiber in interphase cell nuclei.

Attempts have been made to predict the binding structures of the human immunodeficiency virus-1 protease (HIV-1Pr) with various inhibitors within the shortest simulation time consuming. The purpose here is to improve the structural prediction by using statistical approach. We use a combination of molecular docking and non-parametric binomial distribution test considering the combination of binding energy, hydrogen bonding, and hydrophobic-hydrophilic interaction in term of binding residues to select the most probable binding structure. In this study, the binding of HIV-1Pr and two inhibitors: Saquinavir and extracts (3-oxotrirucalla-7, 24-dien-21-oic acid) were investigated. Each inhibitor was positioned in the active site of HIV-1Pr in many different ways using Lamarckian genetic algorithm and then score each orientation by applying a reasonable evaluation function by AutoDock3.0 program. The results from search methods were screened out using non-parametric binomial distribution test and compared with the binding structure from explicit molecular dynamic simulation. Both complexes from statistical selected docking simulation were found to be comparable with those from X-ray diffraction analysis and explicit molecular dynamic simulation structures.