George Barreto Bezerra

Adaptive radius immune algorithm for data clustering

Many algorithms perform data clustering by compressing the original data into a more compact and interpretable representation, which can be more easily inspected for the presence of clusters. This, however, can be a risky alternative, because the simplified representation may contain distortions mainly related to the density information present in the data, which can considerably act on the clustering results. In order to treat this deficiency, this paper proposes an Adaptive Radius Immune Algorithm (ARIA), which is capable of maximally preserving the density information after compression by implementing an antibody adaptive suppression radius that varies inversely with the local density in the space. ARIA is tested with both artificial and real world problems obtaining a better performance than the aiNet algorithm and showing that preserving the density information leads to refined clustering results.

An immunological filter for spam

Spam messages are continually filling email boxes of practically every Web user. To deal with this growing problem, the development of high-performance filters to block those unsolicited messages is strongly required. An Antibody Network, more precisely SRABNET (Supervised Real-Valued Antibody Network), is proposed as an alternative filter to detect spam. The model of the antibody network is generated automatically from the training dataset and evaluated on unseen messages. We validate this approach using a public corpus, called PU1, which has a large collection of encrypted personal e-mail messages containing legitimate messages and spam. Finally, we compared the performance with the well known naive Bayes filter using some performances indexes that will be presented.

Bioinformatics Data Analysis Using an Artificial Immune Network

This work describes a new proposal for gene expression data clustering based on a combination of an immune network, named aiNet, and the minimal spanning tree (MST). The aiNet is an AIS inspired by the immune network theory. Its main role is to perform data compression and to identify portions of the input space representative of a given data set. The output of aiNet is a set of antibodies that represent the data set in a simplified way. The MST is then built on this network, and clusters are determined by using a new method for detecting the inconsistent edges of the tree. An important advantage of this technique over the classical approaches, like hierarchical clustering, is that there is no need of previous knowledge about the number of clusters and their distributions. The hybrid algorithm was first applied to a benchmark data set to demonstrate its validity, and its results were compared with those produced by other approaches from the literature. Using the full yeast S. cerevisiae gene expression data set, it was possible to detect a strong interconnection of the genes, hindering the perception of inconsistencies that may lead to the separation of data into clusters.

An Immune-Evolutionary Algorithm for Multiple Rearrangements of Gene Expression Data

Microarray technologies are employed to simultaneously measure expression levels of thousands of genes. Data obtained from such experiments allow inference of individual gene functions, help to identify genes from specific tissues, to analyze the behavior of gene expression levels under various environmental conditions and under different cell cycle stages, and to identify inappropriately transcribed genes and several genetic diseases, among many other applications. As thousands of genes may be involved in a microarray experiment, computational tools for organizing and providing possible visualizations of the genes and their relationships are crucial to the understanding and analysis of the data. This work proposes an algorithm based on artificial immune systems for organizing gene expression data in order to simultaneously reveal multiple features in large amounts of data. A distinctive property of the proposed algorithm is the ability to provide a diversified set of high-quality rearrangements of the genes, opening up the possibility of identifying various co-regulated genes from representative graphical configurations of the expression levels. This is a very useful approach for biologists, because several co-regulated genes may exist under different conditions.

New Perspectives for the Biclustering Problem

Multimodal optimization algorithms inspired by the immune system are generally characterized by a dynamic control of the population size and by diversity maintenance along the search. One of these proposals, denoted copt-aiNet (artificial immune network for combinatorial optimization), is used to deal with combinatorial problems like the Traveling Salesman Problem (TSP) and other permutation problems. In this paper, the copt-aiNet algorithm is extended and adapted to be applied to an important issue of modern data mining, the biclustering problem. The biclustering approach consists in simultaneously ordering the rows and columns of a given matrix, so that similar elements are grouped together. To illustrate the performance of the proposed method, two bitmap images are scrambled and used as input to the algorithm, and the biclustering procedure tries to restore the original image by grouping the pixels according to the similarity of colors in a neighborhood. Additionally, copt-aiNet is applied to gene expression data clustering, a classical problem of the bioinformatics literature, and its performance is compared with a hierarchical biclustering algorithm.

A hierarchical immune network applied to gene expression data

This paper describes a new proposal for gene expression data analysis. The method used is based on a hierarchical approach to a hybrid algorithm, which is composed of an artificial immune system, named aiNet, and a well known graph theoretic tool, the minimal spanning tree (MST). This algorithm has already proved to be efficient for clustering gene expression data, but its performance may decrease in some specific cases. However, through the use of a hierarchical approach of immune networks it is possible to improve the clustering capability of the hybrid algorithm, such that it becomes more efficient, even when the data set is complex. The proposed methodology is applied to the yeast data and gives important conclusions of the similarity relationships among genes within the data set.

An Immunological Density-Preserving Approach to the Synthesis of RBF Neural Networks for classification

Radial basis function (RBF) neural networks are universal approximators and have been used for a wide range of applications. Aiming at reducing the number of neurons in the hidden layer, for regularization purposes, the center and dispersion of each RBF have to be properly defined by means of competitive learning. Only the output weights will be defined in a supervised manner. One of the drawbacks of such learning methodology, involving unsupervised and supervised learning, is that the centers will be defined so that regions in the input space with a high density of samples tend to be under-represented and those regions with a low density of samples tend to be over-represented. Additionally, few approaches provide a proper and individual indication of the dispersion of each RBF. This paper presents an immune density-preserving algorithm with adaptive radius, called ARIA, to determine the number of centers, their location and the dispersion of each RBF, based only on the available training data set. Considering classification problems, the algorithm to determine the hidden layer is compared to another immune-inspired algorithm called aiNet, K-means and the random choice of centers. The classification accuracy of the final network is compared to another density based approach and a decision tree classifier, C 5.0. The results are reported and analyzed.

Handling Data Sparseness in Gene Network Reconstruction

One of the main problems related to regulatory network reconstruction from expression data concerns the small size and low quality of the available dataset. When trying to infer a model from little information it is necessary to give much more precedence to generalization, rather than specificity, otherwise, any attempt will be fated to overfitting. In this paper we address this issue by focusing on data sparseness and noisy information, and propose a density estimation technique that achieves regularized curves when data is scarce. We first compare the proposed method with the EM algorithm for mixture models on density estimation problems. Next, we apply the method, together with Bayesian networks, on realistic simulations of static gene networks, and compare the obtained results with the standard discrete Bayesian network model. We intend to demonstrate that adopting a discrete approach is not justifiable when only a small amount of information is available.