Antwort: Re: [EMBOSS] using emma: where to put clustalw

[David.Bauer at Schering.de]

Hi Robin,

how long are your 4 sequences ?
I observed that clustalw has problems with nucleotide alignments, if there
are larger differences in sequence length.
So e.g. if a 1 kb sequence is nearly completely contained with high
homology within another 2 kb sequence, the resulting alignment can be very
far from optimal.
If you try to align coding sequences there is a program "tranalign" in
EMBOSS.
You can first align the protein sequences (which usually works better than
a multiple alignment of DNA) and then use this alignment with tranalign to
guide the alignment of the corresponding cDNA.

Hope this helps,
David.



                                                                                                                                 
                      Robin Colgrove                                                                                             
                      <robin at hms.harva                                                                                           
                      rd.edu>                  An:      emboss at embnet.org                                                        
                      Gesendet von:            Kopie:                                                                            
                      owner-emboss at hgm         Thema:   Re: [EMBOSS] using emma: where to put clustalw                           
                      p.mrc.ac.uk                                                                                                
                                                                                                                                 
                                                                                                                                 
                      12.04.2005 21:20                                                                                           
                                                                                                                                 
                                                                                                                                 





Thanks to all for suggestions.
Just putting clustalw in /usr/local/bin did the trick.

Now, I need to figure out why emma/clustalw is giving me such bad
alignments.
Since I only had 4 sequences, I ended up aligning them pairwise with
needle, then pieced together the full alignment in vi, but this is not
going to fly as the number of sequences increases. The online tool I
use for quick alignments (
http://prodes.toulouse.inra.fr/multalin/multalin.html ) does fine, but
the same fasta file sent either to emma or directly to clustalw gives
obviously wrong alignments, even though the nucleotide sequences are
highly homologous.

thanks again

robin



On Apr 12, 2005, at 10:17 AM, Robin Colgrove wrote:

>
> Hello.
>
> I was trying to use emma for a multiple sequence alignment of dna
> sequencing reads, but it complained that it could not find clustalw. I
> could not find any mention of clustalw on the EMBOSS page, so I got a
> copy from the clustalw homepage and -not knowing where to place it-
> tried the /usr/local/share/ EMBOSS/acd directory. That didn't work,
> and emma gives the error:
>
>    EMBOSS An error in ajsys.c at line 398:
> cannot find program 'clustalw'
>
> Looking in the emma.acd and ajsys.c files, I can't find any guidance.
>
> Does anyone know how this is supposed to work?
> Alternatively, is there another good way to do multiple sequence
> alignment?
> Looking ahead, I do not find any obvious way to do contig assembly, a
> la Phrap, or CAP.
>
> thanks
>
> robin colgrove
>