Ozgur Ozturk

PhD Thesis

We present a multi-dimensional indexing approach for fast sequence similarity search in DNA and protein databases. In particular, we propose effective transformations of subsequences into numerical vector domains and build efficient index structures on the transformed vectors. We then define distance functions in the transformed domain and examine properties of these functions. We experimentally compared their (a) approximation quality for k-Nearest Neighbor (k-NN) queries and both (b) pruning… Show more

We present a multi-dimensional indexing approach for fast sequence similarity search in DNA and protein databases. In particular, we propose effective transformations of subsequences into numerical vector domains and build efficient index structures on the transformed vectors. We then define distance functions in the transformed domain and examine properties of these functions. We experimentally compared their (a) approximation quality for k-Nearest Neighbor (k-NN) queries and both (b) pruning ability and (c) approximation quality for ε-range queries. Results for k-NN queries, which we present here, show that our proposed distances FD2 and WD2 (i.e. Frequency and Wavelet Distance functions for 2-grams) perform significantly better than the others. We then develop effective index structures, based on R-trees and scalar quantization, on top of transformed vectors and distance functions. Promising results from the experiments on real biosequence data sets are presented. Show less

In this paper, we present CoMRI, compressed multiresolution index, our system for fast sequence similarity search in DNA sequence databases. We employ virtual bounding rectangle (VBR) concept to build a compressed, grid style index structure. An advantage of grid format over trees is subsequence location information is given by the order of corresponding VBR in the VBR list. Taking advantage of VBRs, our index structure fits into a reasonable size of memory easily. Together with a new optimized… Show more

In this paper, we present CoMRI, compressed multiresolution index, our system for fast sequence similarity search in DNA sequence databases. We employ virtual bounding rectangle (VBR) concept to build a compressed, grid style index structure. An advantage of grid format over trees is subsequence location information is given by the order of corresponding VBR in the VBR list. Taking advantage of VBRs, our index structure fits into a reasonable size of memory easily. Together with a new optimized multiresolution search algorithm, the query speed is improved significantly. Extensive performance evaluations on human chromosome sequence data show that VBRs save 80%-93% index storage size compared to MBRs (minimum bounding rectangles) and new search algorithm prunes almost all unnecessary VBRs which guarantees efficient disk I/O and CPU cost. According to the results of our experiments, the performance of CoMRI is at least 100 times faster than MRS which is another grid index structure introduced very recently. Show less

We present a multi-dimensional indexing approach for fast sequence similarity search in DNA and protein databases. In particular, we propose effective transformations of subsequences into numerical vector domains and build efficient index structures on the transformed vectors. We then define distance functions in the transformed domain and examine properties of these functions. We experimentally compared their (a) approximation quality for k-Nearest Neighbor (k-NN) queries, (b) pruning ability… Show more

We present a multi-dimensional indexing approach for fast sequence similarity search in DNA and protein databases. In particular, we propose effective transformations of subsequences into numerical vector domains and build efficient index structures on the transformed vectors. We then define distance functions in the transformed domain and examine properties of these functions. We experimentally compared their (a) approximation quality for k-Nearest Neighbor (k-NN) queries, (b) pruning ability and (c) approximation quality for E-range queries. Results for k-NN queries, which we present here, show that our proposed distances FD2 and WD2 (i.e. Frequency and Wavelet Distance functions for 2-grams) perform significantly better than the others. We then develop effective index structures, based on R-trees and scalar quantization, on top of transformed vectors and distance functions. Promising results from the experiments on real biosequence data sets are presented. Show less