Catálogo de publicaciones - libros

Many heuristics used for finding biclusters in microarray data use the mean squared residue as a way of evaluating the quality of biclusters. This has led to the discovery of interesting biclusters. Recently it has been proven that the mean squared residue may fail to identify some interesting biclusters. This motivates us to introduce a new measure, called , for assessing the quality of biclusters in microarray data. In order to test the validity of the proposed measure, we include it within an evolutionary algorithm. Experimental results show that the use of this novel measure is effective for finding interesting biclusters, which could not have been discovered with the use of the mean squared residue.

A common way to characterise important and conserved signals in nucleotide sequences, such as transcription factor binding sites, is via the use of so-called sequences or consensus patterns. A well-known example is the so-called “TATA-box” commonly found in eukaryotic core promoters. Such patterns are valuablein that they offer an insight into basic molecular biology processes, and can support reasoning regarding the understanding, design and control of these processes. However it is rare for such patterns to be accurate; instead they represent a very approximate characterisation of the signal under study. At the opposite extreme, we may instead characterise such a signal via a neural network, or a high-order Markov model, and so on. These have better sensitivity and specificity, but are unreadable, and consequently unhelpful for conveying an understanding of the underlying molecular biology processes that could support insight or reasoning. We describe a simple pattern language, called crisp hypermotifs (CHMs), that leads to highly readable patterns that can support understanding and reasoning, yet achieve greater sensitivity and specificity than the commonly used approaches to crisply characterise a signal. We use evolutionary computation to discover high-performance CHMs from data, and we argue that CHMs be used in place of classical consensus motifs, and justify that by presenting examples derived from a large dataset of mammalian core promoters. We provide CHM alternatives to the well-known core promoter TATA-box and Initiator patterns that have better sensitivity and specificity than their classical counterparts.

Although important contributions have been made in recent years within the field of bioprocess model development and validation, in many cases the utility of even relatively good models for process optimization with current state-of-the-art algorithms (mostly offline approaches) is quite low. The main cause for this is that open-loop fermentations do not compensate for the differences observed between model predictions and real variables, whose consequences can lead to quite undesirable consequences. In this work, the performance of two different algorithms belonging to the main groups of (EA) and (DE) is compared in the task of online optimisation of fed-batch fermentation processes. The proposed approach enables to obtain results close to the ones predicted initially by the mathematical models of the process, deals well with the noise in state variables and exhibits properties of graceful degradation. When comparing the optimization algorithms, the seems the best alternative, but its superiority seems to decrease when noisier settings are considered.

Directed search methods and probabilistic approaches have been used as two alternative ways for computational protein design. This paper presents a hybrid methodology that combines features from both approaches. Three estimation of distribution algorithms are applied to the solution of a protein design problem by minimization of contact potentials. The combination of probabilistic models able to represent probabilistic dependencies with the use of information about residues interactions in the protein contact graph is shown to improve the efficiency of search for the problems evaluated.

In human mesenchymal stem cells the envelope surrounding the nucleus, as visualized by the nuclear lamina, has a round and flat shape. The lamina structure is considerably deformed after activation of cell death (apoptosis). The spatial organization of the lamina is the initial structural change found after activation of the apoptotic pathway, therefore can be used as a marker to identify cells activated for apoptosis. Here we investigated whether the spatial changes in lamina spatial organization can be recognized by machine learning algorithms to classify normal and apoptotic cells. Classical machine learning algorithms were applied to classification of 3 image sections of nuclear lamina proteins, taken from normal and apoptotic cells. We found that the Evolutionary-optimized Support Vector Machine (SVM) algorithm succeeded in the classification of normal and apoptotic cells in a highly satisfying result.

This is the first time that cells are classified based on lamina spatial organization using the machine learning approach. We suggest that this approach can be used for diagnostic applications to classify normal and apoptotic cells.

The ability to measure the transcriptional response after a stimulus has drawn much attention to the underlying gene regulatory networks. Here, we evaluate the application of methods to reconstruct gene regulatory networks by applying them to the SOS response of , the budding yeast cell cycle and models. For each network we define an validation network, where each interaction is justified by at least one publication. In addition to the existing methods, we propose a SVD based method (NSS). Overall, most reconstruction methods perform well on data sets, both in terms of topological reconstruction and predictability. For biological data sets the application of reconstruction methods is suitable to predict the expression of genes, whereas the topological reconstruction is only satisfactory with steady-state measurements. Surprisingly, the performance measured on data does not correspond with the performance measured on biological data.

“Quorum sensing” has been described as “the most consequential molecular microbiology story of the last decade” [1][2]. The purpose of this paper is to study the mechanism of quorum sensing, in order to obtain a deeper understanding of how and when this mechanism works. Our study focuses on the use of an Individual-based Modeling (IbM) method to simulate this phenomenon of “cell-to-cell communication” incorporated in bacterial foraging behavior, in both intracellular and population scales. The simulation results show that this IbM approach can reflect the bacterial behaviors and population evolution in time-varying environments, and provide plausible answers to the emerging question regarding to the significance of this phenomenon of bacterial foraging behaviors.

The Bio-basis Function Neural Network (BBFNN) is a novel neural architecture for peptide classification that makes use of amino acid mutation matrices and a similarity function to model protein peptide data without encoding. This study presents an Evolutionary Bio-basis network (EBBN), an extension to the BBFNN that uses a self adapting Evolution Strategy to optimise a problem specific substitution matrix for much improved model performance. The EBBN is assessed against BBFNN and multi layer perceptron (MLP) models using three datasets covering cleavage sites, epitope sites, and glycoprotein linkage sites. The method exhibits statistically significant improvements in performance for two of these sets.