Hi James.
Romain and Andreas have provided some great suggestions. I thought it
worth adding that when using Jalview with web sites and other tools -
the most convenient way to prepare selections for input (e.g. to a tree
calculation) is by creating a new view (Ctrl or Command + T), select all
the columns you want to employ for calculation and use the
'View->Hide->All but selected region' to hide everything that you *do
not* want to employ for the tree calculation (shift+CMD/CTRL + H).
Creating a view minimises the number of additional windows you create
containing the same alignment data, and the input data view can be given
it's own name and archived in a jalview project. Once you have some
results from the other tools, you can then add them the view, and
explore the alignment further via Jalview's subfamily shading methods
(or even the Sequence Harmony service since its designed to predict
functional site variation based on a set of defined subgroups on the
alignment).
If you have any problems exporting alignment data from Jalview or
importing the trees back in to Jalview from RaxML or FastTree, send an
email ! We also now have PHYLIP format input/export support in the
development versions of Jalview which is useful when working with RaxML.
Jim.
PS. Just to expand on Romain's comment about accuracy: Jalview's tree
algorithms are rigorous but relatively primitive, and not considered
appropriate for general phylogenetic analysis tasks. Also, Jalview
doesn't provide support for model selection (picking the right model to
calculate the intersequence distances from the alignment) or
bootstrapping (identifying the statistically significant branches in the
tree). FastTree and RaxML both employ heuristic maximum likelihood
searches to produce an accurate tree more quickly and includes
approximate support calculations.
On 06/11/2014 09:35, rstuder wrote:
Yes, trees in Jalview are calculated only based on marked positions.
For accuracy, I would not use phylogenetic tools from Jalview.
I would rather do the following:
1) Select the positions in Jalview.
2) Copy them.
3) Paste them as new alignment.
4) Save the alignment in a new file.
And then I use FastTree or RAxML (or even PhyML if you have access to
good computer cluster).
Romain
On 05/11/2014 20:18, James Starlight wrote:
Hi Romain,
thank you very much for the explanation!
I've already used TrimAl as the part of the Phylemon2 server and
found it very useful :-)
Regarding calculating Trees in JalView using subset of the residues:
as I noticed the tree are calculated just in case when the
positions of ligand-contacting residues are marked by red color in
the top, aren't it? Is it possible in addition to check on what
exactly subset trees has been calculated based on the output results?
Finally regarding accuracy of the calculation of the trees- what
method should produce best results for the alignment consisted of
several hundred of sequences?
James
2014-11-05 16:59 GMT+01:00 rstuder <[email protected]
<mailto:[email protected]>>:
Dear James,
For aligning the sequences, I would recommend MAFFT (L-INS-i) or
Clustal-Omega:
http://mafft.cbrc.jp/alignment/software/
http://www.clustal.org/omega/
Then, you can use Jalview to select manually the ligand-binding
domain in your alignment.
You can also use TrimAl to select only well aligned position that
corresponding to phylogenetic signal.
http://trimal.cgenomics.org/
For producing the tree on very big alignments, I would recommend
FastTree. It produces quite good results and is very easy to use:
http://www.microbesonline.org/fasttree/
There is also RAxML which is developed for big alignments:
http://sco.h-its.org/exelixis/web/software/raxml/index.html
Best regards,
Romain
On 05/11/2014 15:06, James Starlight wrote:
Dear JalView users!
I need to perform large-scale phylogenetic analysis of big
dataset of GPCR sequences selecting as the input only residues
involved in the ligand-binding site of those receptors (taken
from the structural data) from the input multiple-sequence
alignment. I wounder what method for phylogenetic trees
calculation will be best (neighbourhood joining or
pairs-distance calculations) for my task as well as how to make
selection of the selecting residues properly (previously I've
done it by cntrl-left click on the bottom of the alignment
marking corresponded zone by red colour). On what additional
details should I paid my attention during such calculations in
case when I'm dealing with a very big number of sequences?
Thank you for the help,
James
_______________________________________________
Jalview-discuss mailing list
[email protected] <mailto:[email protected]>
http://www.compbio.dundee.ac.uk/mailman/listinfo/jalview-discuss
--
Romain Studer
EMBL-EBI
Wellcome Trust Genome Campus
Hinxton
Cambridgeshire
CB10 1SD, UK
Tel:+44 (0)1223 492 547 <tel:%2B44%20%280%291223%20492%20547>
Twitter: @RomainStuder
--
Romain Studer
EMBL-EBI
Wellcome Trust Genome Campus
Hinxton
Cambridgeshire
CB10 1SD, UK
Tel: +44 (0)1223 492 547
Twitter: @RomainStuder
_______________________________________________
Jalview-discuss mailing list
[email protected]
http://www.compbio.dundee.ac.uk/mailman/listinfo/jalview-discuss
_______________________________________________
Jalview-discuss mailing list
[email protected]
http://www.compbio.dundee.ac.uk/mailman/listinfo/jalview-discuss
