[BiO BB] Observation: multiple sequence alignment affected by the input sequence order

[Iain Wallace]

Hi,

I find this behavior very strange, as the programmes are designed not to
exhibit this behavior.
The first step in must alignment programmes is an all against all
comparison, from which a tree is built. This tree is then used to determine
the order in which the sequences are aligned. There is no dependence on
input order in any of the alignment methods mentioned.

There are a few methods that can be used to compare alignments (and to make
sure that they are identical when only the ordering is changed), such as
aln_compare from Cedric Notredame, Q_score from Bob Edgar (
http://www.drive5.com/) or veralign from Jaap Heringa (online server,
http://zeus.cs.vu.nl/programs/veralignwww/)

I would recommend that you redo your alignment using any of the programmes
you mentioned, and then change the input order and then compare the two
alignments....FYI clustal has an option to output the alignment in the order
that the sequences were aligned, and this shouldn't change regardless of the
input order.

Hope this helps

Iain



On 8/16/07, Hongyu Zhang <forward at hongyu.org> wrote:
>
> Dear all,
>
> I've observed that several multiple sequence alignment programs, including
> ProbCons, ClustalW and Musle, all share the same behavior, i.e., given a
> group of sequences in FASTA format as the input, if I change the order of
> the sequences in the input file, the results generated by those programs
> will change as well. It's not just the sequence order that will change, but
> also the amino acid matches.
>
> I think it's a little counter-intuitive because one would expect the
> opposite. Is there a program that can output a stable alignment independent
> of the input sequence order? Thanks!
>
> _______________________________________________
> General Forum at Bioinformatics.Org -
> BiO_Bulletin_Board at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bio_bulletin_board
>