In this paper, we study the problem of sequence similarity search. We incorporate vector transformations and apply DFT (Discrete Fourier Transformation) and DWT (Discrete Wavelet Transformation, Haar) dimensionality reduction techniques to reduce the search space/time of sequence similarity range queries. Our empirical results on a number of Prokaryote and Eukaryote DNA contig databases demonstrate up to 50-fold filtration ratio reduction of the search space and up to 13 times faster filtration. The proposed transformation techniques may easily be integrated as a pre-processing phase on top of current similarity search heuristics/techniques such as BLAST, PatternHunter, FastA and QUASAR to efficiently prune non-relevant sequences. We study the precision of applying dimensionality reduction techniques for faster and more efficient range query searches and discuss the imposed trade-offs.

A novel data mining approach for similarity search and knowledge discovery in protein structure databases is proposed. PADS (Protein structure Alignment by Directional shape Signatures) incorporates the three dimensional coordinates of the main atoms of each amino acid and extracts a geometrical shape signature along with the direction of each amino acid. As a result, each protein structure is presented by a series of multidimensional feature vectors representing local geometry, shape, direction, and biological properties of its amino acid molecules. Furthermore, a distance matrix is calculated and is incorporated into a local alignment dynamic programming algorithm to find the similar portions of two given protein structures followed by a sequence alignment step for more efficient filtration. The optimal superimposition of the detected similar regions is used to assess the quality of the results. The proposed algorithm is fast and accurate and hence could be used for analysis and knowledge discovery in large protein structures. The method has been compared with the results from CE, DALI, and CTSS using a representative sample of PDB structures. Several new structures not detected by other methods are detected.

Skyline queries help users make intelligent decisions over complex data, where different and often conflicting criteria are considered. Current skyline computation methods are restricted to centralized query processors, limiting scalability and imposing a single point of failure. In this paper, we address the problem of parallelizing skyline query execution over a large number of machines by leveraging content-based data partitioning. We present a novel distributed skyline query processing algorithm (DSL) that discovers skyline points progressively. We propose two mechanisms, recursive region partitioning and dynamic region encoding, to enforce a partial order on query propagation in order to pipeline query execution. Our analysis shows that DSL is optimal in terms of the total number of local query invocations across all machines. In addition, simulations and measurements of a deployed system show that our system load balances communication and processing costs across cluster machines, providing incremental scalability and significant performance improvement over alternative distribution mechanisms.