A linear space algorithm for computing maximal common subsequences
Communications of the ACM
Proceedings of the 5th International Conference on Intelligent Systems for Molecular Biology
OASIS: an online and accurate technique for local-alignment searches on biological sequences
VLDB '03 Proceedings of the 29th international conference on Very large data bases - Volume 29
| Hi-index | 0.00 |
There is a pressing need to align growing set of expressed sequence tags (ESTs)to newly sequenced human genome that is still frequently revised, for providing biologists and medical scientists with fresh information. The problem is, however, complicated by the exon/intron structure of eucaryotic genes, misread nucleotides in ESTs, and millions of repeptive sequences in genomic sequences. Indeed, to solve this, algorithms that use dynamic programming have been proposed, in which space complexity is O(N) and time complexity is O(MN) for a genomic sequence of length M and an EST of length N, but in reality, these algorithms require an enormous amount of processing time. In an effort to improve the computational efficiency of these classical DP algorithms, we develop software that fully utilizes the lookup-table that stores the position at which each short subsequence occurs in the genomic sequence for allowing the efficient detection of the start- and endpoints of an EST within a given DNA sequence, and subsequently, the prompt identification of exons and introns. In addition, high sensitivity and accuracy must be achieved by calculating locations of all spliced sites correctly for more ESTs while retaining high computational efficiency. This goal is hard to accomplish in practice, owing to misread nucleotides in ESTs and repeptive sequences in the genome, but we present a couple of heuristics effective in settling this issue. Experimental results have confirmed that our technique improves the overall computation time by orders of magnitude compared with common tools such as sim4 and BLAT, and attains high sensitivity and accuracy against datasets of clean and documented genes at the same time. Consequently, our software is able to align about three millions of ESTs to a draft genome in less than one day, and all the information is available through the WWW at http://grl.gi.k.u-tokyo.ac.jp/.