Multiple Sequence Alignment Using SAGA: Investigating the Effects of Operator Scheduling, Population Seeding, and Crossover Operators

Abstract

Multiple sequence alignment (MSA) is a fundamental problem of great importance in molecular biology. In this study, we investigated several aspects of SAGA, a well-known evolutionary algorithm (EA) for solving MSA problems. The SAGA algorithm is important because it represents a successful attempt at applying EAs to MSA and since it is the first EA to use operator scheduling on this problem. However, it is largely undocumented which elements of SAGA are vital to its performance. An important finding in this study is that operator scheduling does not improve the performance of SAGA compared to a uniform selection of operators. Furthermore, the experiments show that seeding SAGA with a ClustalW-derived alignment allows the algorithm to discover alignments of higher quality compared to the traditional initialization scheme with randomly generated alignments. Finally, the experimental results indicate that SAGA’s performance is largely unaffected when the crossover operators are disabled. Thus, the major determinant of SAGA’s success seems to be the mutation operators and the scoring functions used.