Efficient Index-Based Methods For Processing Large Biological Databases.

Over the last few decades, advances in life sciences have generated a vast amount of biological data. To cope with the rapid increase in data volume, there is a pressing need for efficient computational methods to query large biological datasets. This thesis develops efficient and scalable querying methods for biological data. For an efficient sequence database search, we developed two q-gram index based algorithms, miBLAST and ProbeMatch. miBLAST is designed to expedite batch identification of statistically significant sequence alignments. ProbeMatch is designed for identifying sequence alignments based on a k-mismatch model. For an efficient protein structure database search, we also developed a multi-dimensional index based algorithm method called proCC, an automatic and efficient classification framework. All these algorithms result in substantial performance improvements over existing methods. When designing index-based methods, the right choice of indexing methods is essential. In addition to developing index-based methods for biological applications, we also investigated an essential database problem that reexamines the state-of-the-art indexing methods by experimental evaluation. Our experimental study provides a valuable insight for choosing the right indexing method and also motivates a careful consideration of index structures when designing index-based methods. In the long run, index-based methods can lead to new and more efficient algorithms for querying and mining biological datasets. The examples above, which include query processing on biological sequence and geometrical structure datasets, employ index-based methods very effectively. While the database research community has long recognized the need for index-based query processing algorithms, the bioinformatics community has been slow to adopt such algorithms. However, since many biological datasets are growing very rapidly, database-style index-based algorithms are likely to play a crucial role in modern bioinformatics methods. The work proposed in this thesis lays the foundation for such methods.