Giuseppe Pigola

Enhancing density-based clustering: Parameter reduction and outlier detection

Clustering is a widely used unsupervised data mining technique. It allows to identify structures in collections of objects by grouping them into classes, named clusters, in such a way that similarity of objects within any cluster is maximized and similarity of objects belonging to different clusters is minimized. In density-based clustering, a cluster is defined as a connected dense component and grows in the direction driven by the density. The basic structure of density-based clustering presents some common drawbacks: (i) parameters have to be set; (ii) the behavior of the algorithm is sensitive to the density of the starting object; and (iii) adjacent clusters of different densities could not be properly identified. In this paper, we address all the above problems. Our method, based on the concept of space stratification, efficiently identifies the different densities in the dataset and, accordingly, ranks the objects of the original space. Next, it exploits such a knowledge by projecting the original data into a space with one more dimension. It performs a density based clustering taking into account the reverse-nearest-neighbor of the objects. Our method also reduces the number of input parameters by giving a guideline to set them in a suitable way. Experimental results indicate that our algorithm is able to deal with clusters of different densities and outperforms the most popular algorithms DBSCAN and OPTICS in all the standard benchmark datasets.

MIDClass: microarray data classification by association rules and gene expression intervals.

We present a new classification method for expression profiling data, called MIDClass (Microarray Interval Discriminant CLASSifier), based on association rules. It classifies expressions profiles exploiting the idea that the transcript expression intervals better discriminate subtypes in the same class. A wide experimental analysis shows the effectiveness of MIDClass compared to the most prominent classification approaches.

Anticlustal: multiple sequence alignment by antipole clustering and linear approximate 1-median computation

In this paper we present a new multiple sequence alignment (MSA) algorithm called AntiClustAl. The method makes use of the commonly used idea of aligning homologous sequences belonging to classes generated by some clustering algorithm, and then continue the alignment process in a bottom-up way along a suitable tree structure. The final result is then read at the root of the tree. Multiple sequence alignment in each cluster makes use of the progressive alignment with the 1-median (center) of the cluster. The 1-median of set S of sequences is the element of S which minimizes the average distance from any other sequence in S. Its exact computation requires quadratic time. The basic idea of our proposed algorithm is to make use of a simple and natural algorithmic technique based on randomised tournaments which has been successfully applied to large size search problems in general metric spaces. In particular a clustering algorithm called antipole tree and an approximate linear 1-median computation are used. Our algorithm compared with Clustal W, a widely used tool to MSA, shows a better running time results with fully comparable alignment quality. A successful biological application showing high amino acid conservation during evolution of Xenopus laevis SOD2 is also cited.

Sequence similarity is more relevant than species specificity in probabilistic backtranslation

BackgroundBacktranslation is the process of decoding a sequence of amino acids into the corresponding codons. All synthetic gene design systems include a backtranslation module. The degeneracy of the genetic code makes backtranslation potentially ambiguous since most amino acids are encoded by multiple codons. The common approach to overcome this difficulty is based on imitation of codon usage within the target species.ResultsThis paper describes EasyBack, a new parameter-free, fully-automated software for backtranslation using Hidden Markov Models. EasyBack is not based on imitation of codon usage within the target species, but instead uses a sequence-similarity criterion. The model is trained with a set of proteins with known cDNA coding sequences, constructed from the input protein by querying the NCBI databases with BLAST. Unlike existing software, the proposed method allows the quality of prediction to be estimated. When tested on a group of proteins that show different degrees of sequence conservation, EasyBack outperforms other published methods in terms of precision.ConclusionThe prediction quality of a protein backtranslation methis markedly increased by replacing the criterion of most used codon in the same species with a Hidden Markov Model trained with a set of most similar sequences from all species. Moreover, the proposed method allows the quality of prediction to be estimated probabilistically.

FAST CLUSTERING AND MINIMUM WEIGHT MATCHING ALGORITHMS FOR VERY LARGE MOBILE BACKBONE WIRELESS NETWORKS

Mobile Backbone Wireless Networks (MBWN) [10] are wireless networks in which the base stations are mobile. Our strategy is the following: mobile nodes are dynamically grouped into clusters of bounded radius. In the very large wireless networks we deal with we deal with, several hundreds of clusters may be generated. Clustering makes use of a two dimensional Euclidean version of the Antipole Tree data structure [5]. This very effective structure was originally designed for finite sets of points in an arbitrary metric space to support efficient range searching. It requires only a linear number of pair-wise distance calculations among nodes. Mobile Base Stations occupy an approximate centroid of the clusters and are moved according to a fast practical bipartite matching algorithm which tries to minimize both total and maximum distance. We show that the best known computational geometry algorithms [1] become infeasible for our application when a high number of mobile base stations is required. On the other hand our proposed 8% average error solution requires O(k log k) running time instead of the approximatively O(k2) exact algorithm [1]. Communication among nodes is realized by a Clusterhead Gateway Switching Routing (CGSR) protocol [15] where the mobile base stations are organized in a suitable network. Other efficient clustering algorithms [11, 17] may be used instead of the Antipole Tree. However the nice hierarchical structure of the Antipole Tree makes it applicable to other types of mobile wireless (Ad-Hoc) and wired networks but this will be subject of future work.

Obstacles constrained group mobility models in event-driven wireless networks with movable base stations

In this paper, we propose a protocol for dynamic reconfiguration of ad-hoc wireless networks with movable base stations in presence of obstacles. Hosts are assigned to base stations according to a probabilistic throughput function based on both the quality of the signal and the base station load. In order to optimize space coverage, base stations cluster hosts using a distributed clustering algorithm. Obstacles may interfere with transmission and obstruct base stations and hosts movement. To overcome this problem, we perform base stations repositioning making use of a motion planning algorithm on the visibility graph based on an extension of the bottleneck matching technique. We implemented the protocol on top of the NS2 simulator as an extension of the AODV. We tested it using both Random Way Point and Reference Point Group mobility models properly adapted to deal with obstacles. Experimental analysis shows that the protocol ensures the total space coverage together with a good throughput on the realistic model (Reference Point Group) outperforming both the standard AODV and DSR.

Elucidating the Role of microRNAs in Cancer Through Data Mining Techniques

microRNAs (miRNAs) have been shown to play a crucial role in the most important biological processes and their dysregulation has been connected to a variety of diseases, including cancer. The number of computational tools for the analysis of miRNA related data is continuously increasing. They range from simple look-up resources to more sophisticated tools for functional analysis of miRNAs. These systems may help to investigate the role of miRNAs in key biological processes and their involvement in diseases. The ultimate goal is to allow the development of regulatory models describing complex processes and the effects of their dysregulation.Here we review the most important and recent methods for the analysis of miRNA expression profiles and the tools available on the web for target prediction and functional analysis of miRNAs.Particular emphasis is given to the integration of heterogeneous data, including target predictions and expression profiles, which can be used to infer miRNA/phenotype associations and for the generation of network models of miRNA function.