Search

Computational methods on molecular sequence data are at the heart of computational molecular biology. Identification of known or unknown DNA and RNA motifs or regions involved in various biological processes such as initiation of transcription, gene expression and translation, or the discovery of various types of repeats are some of the applications of major concern. An accurate identification and localization of such elements will allow biologists to perform deeper studies of the structure, function and evolution of genomes. This requires the development of faster and more complex mathematical models and computer algorithms. In this work we discuss current techniques to cope with string problems in molecular sequence data. We focus on Weighted Sequences and Sequences with "don't care characters", explaining the open problems and their relevance to biological applications. [ABSTRACT FROM AUTHOR]

In this paper we introduce the Weighted Suffix Tree, an efficient data structure for computing string regularities in weighted sequences of molecular data. Molecular Weighted Sequences can model important biological processes such as the DNA Assembly Process or the DNA-Protein Binding Process. Thus pattern matching or identification of repeated patterns, in biological weighted sequences is a very important procedure in the translation of gene expression and regulation. We present time and space efficient algorithms for constructing the weighted suffix tree and some applications of the proposed data structure to problems taken from the Molecular Biology area such as pattern matching, repeats discovery, discovery of the longest common subsequence of two weighted sequences and computation of covers. [ABSTRACT FROM AUTHOR]

Many alternatives have been proposed for the mining of association rules involving rare but 'interesting' itemsets in a dataset where there also exist highly frequent itemsets. Nevertheless, all the approaches thus far suggested that we knew which those interesting itemsets are, as well as which is the right support value for them. None of the approaches proposed a way of automatically discovering such items. In this work we introduce the notion of locally frequent itemsets and support their existence as the biggest and most frequently appearing category of rare but interesting itemsets especially at commercial applications, based on the opinion of field experts. Subsequently we propose two algorithms for finding and handling these itemsets. The main idea is to divide the database into partitions according to the problem needs and besides searching for itemsets which are frequent in the whole database to search also for itemsets which are frequent if considered within these partitions. Our approach proves very effective and also very efficient as compared to the traditional algorithms both in synthetic and real data. [ABSTRACT FROM AUTHOR]