A generic motif discovery algorithm for sequential data

Abstract

Gemoda is freely available at http://web.mit.edu/bamel/gemoda

Motif-Based Classification of Time Series with Bayesian Networks and SVMs

“…Jensen at al. [13] and Ferreira at al. [7] also use clustering, however in a more local fashion: they do not cluster the whole sequences, but subsequences of them.…”

“…Predefining a (minimal) length L for motifs, scanning the database, and enumerating (almost) all the subsequences of the given lenght L is common in the biological domain [31,7,13,21]. However this may not be efficient enough, especially if complex motifs with gaps and/or taxonomical wildcards are to be discovered (see Fig.…”

Mining Longest Common Subsequence and other Related Patterns using DNA Operations

“…Step 2 generate DNA 0 , the possible combinations of 0's and DNA 1 , the possible combinations of 1's [18], depending on the level_number. Let the number_of_nodes be the variable, which stores the total number of elements present in DNA 0 or DNA 1 respectively.…”

“…To overcome the difficulty of alignment problem a modified Position Weight Matrices (PWM) [4] can be used to focus on the positions of the patterns in the sequences. Various ways of building a PWM have been carried out, some of them are found in [10,17,18,33,36]. A number of pattern discovery algorithms have been steadily appearing in the literature [1, 4,7,12,13,14,15,19,25,27].…”

“…The majority of the tools can be found at the extreme ends of the spectrum with tools that exhaustively enumerate regular expressions at one end and probabilistic tools, based on Position Weight Matrices(PWMs), at the other. This partitioning of tools is due to a computational trade-off: more descriptive motif representations such as PWMs frequently make exhaustive searches computationally infeasible [18]. The definition of the search problem, especially the formulation of objective functions, leaves space for substantial improvement in the performance of the motif discovery tool [20].…”