Sequence alignment and promoter search algorithms

Abstract

Sequence alignment is a basic method in computational approach to analyze biological sequences. The importance and impact of the improvements in alignments is so great in bioinformatics that more fundamental and basic researches are required. The improvement of both pairwise alignment and multialignment algorithms have been tried in this thesis. An application algorithm using automata in biological sequence pattern recognition and a new filtering algorithm used in contig assembly program have been developed. And a promoter search algorithm has been developed to implement a practical application system. A new pairwise algorithm using linear space is faster than the commonly used Myers and Miller’s algorithm and it is expected that the algorithm could replace it, though in theoretical aspect, more accurate investigation should be done to analyze algorithmic complexity. A new multialignment algorithm uses very different approach from conventional progressive multialignment algorithms. It also uses its own unique data structures. Though the effect could be proved through effective implementation, it is expected that the algorithm could replace the current algorithms because it will overcome many defects of the conventional algorithms. Search algorithms of biological sequences patterns using automata showed good running time efficiency compared with those of general search methods. Such application will be useful for motif pattern recognition and other exact pattern searches. A new filter algorithm using dimer frequency increased the total efficiency of contig assembly program by rapidly and selectively filtering non-overlapping fragments with a high fidelity. A promoter search algorithm was implemented to a practical program running on UNIX. The usefulness and promise were proved through a case study with bacterial classIII promoters. Two genes b2737 and ppdA captured by the promoter search program were proved to be related with $\acute{o}^F$ control, through ...