This is the method described in Barton and Sternberg (1987b). The sequences are aligned progressively in one order. If the sequences are all clearly similar to each other - say > 50%identity, then the actual order will make very little difference to the final alignment. In general however, it is better to establish the order by first performing all pairwise comparisons for the sequences as described in the previous section, then using program ORDER to define an order (described below) before proceeding to multiple alignment.
The ORDER program generates an order_file. This specifies the order in which the sequences present in the seq_file are to be aligned. A typical command file for multiple alignment is given in globin_mult.com and shown below.
Commands Explanation number -------- ------------------ output_file=globin_mult.out 1. mode=multiple 2. matrix=file=ampsdir:md.mat 3. gap_penalty=8.0 4. constant = 8 5. consplot=mz 6. print_vertical= 7. seq_file=globin.seq 8. order_file=globin.ord 9.
Optionally a process of iteration can be performed. Once all the sequences have been added to the alignment, the first sequence can be realigned with the ALIGNMENT of the sequences 2-N. Then the second sequence is aligned with 1,3-N and so on. To specify that iteration should be performed include the command 'iterations=int' where int is an integer greater than 0. In general iteration values of 3 can refine the alignments slightly and are of greatest use if there are relatively few sequences to be aligned (say <10).
The result of a run using this command file is illustrated in the file globin_mult.out.