Alignment and comparison of multiple related DNA sequences can be improved by consideration of known patterns of mutation.

The rate at which bases are replaced in mammalian DNA is influenced by the sequence involved. I have modified a multiple sequence alignment and comparison program to take account of the observed patterns of base replacement in mammals, and used the program to analyse 100 human Alu sequences. The results show that using a sequence matching matrix based on the actual rate of base replacement, as opposed to arbitrary base matching matrix, gives significant differences in the way that sequences are aligned and compared with each other. Use of the observed matching matrix gives significantly better discrimination between sub-families of Alu sequences than conventional methods, and shows that comparison systems which take account of the mechanism of mutation give more biologically realistic results.