[ccp4bb]: Summary: Structure-based sequence alignment

["fan zhang" <zhangfan1980@hotmail.com>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

Hi everyone,

Thankyou very much for your good suggestions. I believe your suggestions 
will benefit those who are preparing a structure-based sequence alignment.
Here is a summary:

Xue-yuan Pei:
Programs Fugue, Joy and Homstrad will do the job for you.

Jens Kaiser:
a working (but tedious :-)) way is using top3d from ccp4 with the two 
coordinate-files. regions corresponding to each other are printed out; 
regions wich have no structural homology (in terms of 3d-coordinates) are 
not printed. you can assemble the parts on top of your two 
complete-sequences in gcg ( or similar program). it takes some time, but you 
will have a 'real' structure based alignement.

Bart Hazes:
If you have structures of both (or more) proteins involved then just 
superimpose them and align the structurally equivalent residues. Deciding 
when to call two residues structurally equivalent can be a bit arbitrary. I 
normally let O's lsq_refi sort out the equivalent bits. Maybe not perfect, 
but no user-bias and a lot easier/faster then eyeballing the structures.
If you only have one structure and want to align other related sequences 
with it you can help normal alignment programs, such as ClustalX, by 
providing it with the location of secondary structure elements. You can also 
align sequences to structure using threading, but the alignments are often 
not very accurate. The THREADER program can align based on both 
physicochemical positions and sequence homology. Maybe with proper weighting 
this is useful. I would personally use standard sequence-only alignments and 
then analyse the conserved features wrt the structure. Do conserved 
hydrophobic positions correspond to hydrophobic core residues, does 
conservation of small residues correspond to a cramped space etc. Based on 
such information you can often spot errors in alignments of highly divergent 
sequences where sequence alone cannot distinguish between alternate 
alignment possibilities.
If they don't occupy structurally equivalent positions than you shouldn't 
normally align them even if they have similar physicochemical properties. If 
the sequence-suggested alignment is clearly at odds with the observed 
structural alignment then I would suggest the sequence evidence has to be 
pretty strong to overrule the structural information. I guess it is possible 
to have an insertion in a helix leading to a shift of the subsequent 
residues. I only know this happening when human beings make mutants, but 
maybe you can find examples in nature. In this case you'll have to decide if 
the sequence alignment is supposed to show evolutionary equivalence (the 
normal assumption) or structural equivalence.

V.S.Gowri:
To my knowledge, u can use STAMP (a rigid body superposition program) by 
Russell and Barton.

Alexander.Pautsch:
you would need to do superimpose the relevant structures (e.g. with lsqkab) 
and then go ahead and (manually) adjust the sequence alignment so that 
corresponding residues match each other. I think lsqman from the Uppsala 
software factory has a command to do this http://alpha2.bmc.uu.se/usf/

Jason Maynes:
The most accurate way for a structure-based alignment is to run your two 
pdb's through the DALI server - http://www2.ebi.ac.uk/dali/. There is also a 
program called mammoth that works well, but is tough to get your hands on.

Gerard J.  Kleywegt:
if you have LSQMAN available, this is a breeze. see the example at 
http://xray.bmc.uu.se/usf/lsqman_man.html#H6 (and get the macro from 
ftp://xray.bmc.uu.se/pub/gerard/omac/align.lsqmac). this will give you the 
structure-based sequence alignment of the structurally conserved parts. 
parts of both sequences that are not conserved in the structure will not be 
aligned by sequence either

Petrus Hendrik Zwart:
Stralign might be an option :
http://www.hgc.ims.u-tokyo.ac.jp/service/tooldoc/stralign/intro.html ( you 
might want to take  a look at this as well: 
http://www.prosci.uci.edu/Articles/Vol9/issue11/0235/0235.html )

Frank Vondelft:
http://www-cryst.bioc.cam.ac.uk/~joy Or more automatic:
http://www-cryst.bioc.cam.ac.uk/~jiye/evoltrace/evoltrace.html which will, 
amongst others, produce the JOY alignment you're after.  And a very handy 
evolutionary tree, so you can include non-structure sequences, which are 
often handy to have aligned.

Good lucks for all who responded !!


cheers.

Zhangfan

Life School of Peking University
P. R. China



_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com