Parallel exact time series motif discovery
Euro-Par'10 Proceedings of the 16th international Euro-Par conference on Parallel processing: Part II
A disk-aware algorithm for time series motif discovery
Data Mining and Knowledge Discovery
| Hi-index | 0.00 |
Life science research labs today manage increasing volumes of sequence data. Much of the data management and querying today is accomplished procedurally using Perl, Python, or Java programs that integrate data from different sources and query tools. The dangers of this procedural approach are well known to the database community (a) severe limitations on the ability to rapidly express queries and (b) inefficient query plans due to the lack of sophisticated optimization tools. This situation is likely to get worse with advances in high-throughput technologies that make it easier to quickly produce vast amounts of sequence data. The need for a declarative and efficient system to manage and query biological sequence data is urgent. To address this need, we designed the Periscope/SQ system. Periscope/SQ extends current relational systems to enable sophisticated queries on sequence data and can optimize and execute these queries efficiently. This thesis describes the problems that need to be solved to make it possible to build the Periscope/SQ system. First, we describe the algebraic framework which forms the backbone of Periscope/SQ. Second, we describe algorithms to construct large scale suffix tree indexes for efficiently answering sequence queries. Third, we describe techniques for selectivity estimation and optimization in the context of queries over biological sequences. Next, we demonstrate how some of the techniques developed for Periscope/SQ can be applied to produce a powerful mining algorithm that we call FLAME. Finally, we describe GeneFinder, a biological application built on top of Periscope/SQ. GeneFinder is currently being used to predict the targets of transcription factors. Today, genomic and proteomic sequences are the most abundantly available source of high-quality biological data. By making it possible to declaratively and efficiently query vast amount of sequence data, Periscope/SQ opens the door to vast improvements in the pace of bioinformatics research.