Genetic algorithm approaches for efficient multiple molecular sequence alignment

Abstract

Multiple biomolecular sequence alignment is among the most important and challenging tasks in computational biology. Current approaches are characterized by great complexity in computational time. The complexity has limited the use of the approaches in many practical applications.

In this research, new approaches based on a genetic algorithm have been developed for multiple biomolecular sequence alignment. Their major strengths are very high efficiency and good alignment quality. Experiments using real data sets have shown that the average computing time of these approaches is one to three orders of magnitude lower than that of a most widely used program while the qualities are very similar.

The key component of these approaches is an enhanced genetic algorithm. Genetic algorithms are a set of stochastic algorithms for efficient and robust search. The basic idea of this approach is the conversion of multiple sequence alignment into a search problem. The conversion enables us to apply a genetic algorithm for efficient identification of matches between multiple sequences.

Three methods, two of them based on dynamic programming, have been developed to handle mismatches. The combination of the genetic algorithm and these methods may produce high quality alignments in an efficient manner.

In this thesis, the theoretical fundamentals of the approaches are discussed, the procedures of the enhanced genetic algorithm as well as the three methods are presented and analyzed, and the experimental results are described and compared with the results obtained by using a most widely used multiple molecular sequence alignment program.