Hi Pedro, I suggest that you use the PROMALS3D because "PROMALS3D constructs alignments for multiple protein sequences and/or structures using information from sequence database searches, secondary structure prediction, available homologs with 3D structures and user-defined constraints"

- http://prodata.swmed.edu/promals3d/promals3d.php

Aligning a protein sequence on already aligned protein sequences does not appears valid.Do a fresh multi sequence alignment including your sequence of interest.If you want to see structural conservation then you can build homology model of your protein and then align all proteins to see their sequence as well structural alignment.

EsPript2.0 assign secondary structure to multiple aligned protein sequences.

I agree that you should not force one sequence into an structure alignment. If you still want to do it, muscle have an option that allow you to pass a profile and your sequence (manual=>http://www.drive5.com/muscle/muscle_userguide3.8.html#_Toc260497020):

'''

2.10 Adding sequences to an existing alignment

To add a sequence to an existing alignment that you wish to keep intact, use profile-profile alignment with the new sequence as a profile. For example, if you have an existing alignment existing_aln.afa and want to add a new sequence in new_seq.fa, use the following commands:

muscle -profile -in1 existing_aln.afa -in2 new_seq.fa -out combined.afa

'''

Also, you can try to use servers or software that give you an alignment based on structure templates. I thought 3Dcoffee (one of the options of t-coffee) did that. Since you already try it, now use promals (http://prodata.swmed.edu/promals3d/promals3d.php) or similar.

Hope this is of help.

Cheers!

T-Coffee should be able to do this correctly, but you have to use the right options. One possibility is to input your existing alignment as a profile and then align your sequence to this profile. Another way would be to use the mode expresso to do a new structural alignment for all sequences for which structures are availlable (this is checked automatically by expresso).

http://tcoffee.crg.cat/apps/tcoffee/index.html

Thank you all for the answers. One can certainly align a sequence to a profile and that is good for detecting remote homologues, protein domains etc. I like and I use tcoffee and muscle but they did not work well this time, not even t-coffee with expresso. Aligning sequences of known structures is straight-forward but it is not that easy when you have a add a new sequence that it is distant relative. To be more specific I am aligning a family of proteins that are structurally related and yet sequence identity can be as low as 15%. T-coffee has a choice in which you can use the secondary structure to guide the alignment. I did not try that because I need to figure out the format of input files. Any help with that will be welcome.

Hi, so 15% is quite low, it's not even sure your proteins are related, I suppose that the identity is mainly due to the cysteine residues ? I assume that you have 3 or more disulphide bridges ? What you could do is create an HMM profile from your structural alignment and run it on your sequence. If you have an E-value greater than 10-3 then it's more than tricky. I don't know if you tried already to use HMMer ? I add similar problem with poorly conserved cysteine rich small protein and it helped to identify and classify a few of them however not all as the structure is mainly constrained by the cysteine residues.

Cedrik that's where I am. Cys are important in my proteins but not a single residue is conserved 100% in the structural alignment. I tried HMMer but I was not happy with the result. Just a small portion of the protein got aligned. I am doing modeling and checking disulphide connectivity to asses, more o less, the quality of the alignment. Not sure this is right.

If you make your multiple sequence alignment into clustalw format (.aln) and remove the top two lines I think, then you can make a PSSM using option -B in psi-blast (blastpgp). I have not tried that myself but give it a go - so something like this: blastpgp -i query.txt -B mult.aln -d mult.fasta -C output.pssm

Hi Pedro, I suggest that you use the PROMALS3D because "PROMALS3D constructs alignments for multiple protein sequences and/or structures using information from sequence database searches, secondary structure prediction, available homologs with 3D structures and user-defined constraints"

- http://prodata.swmed.edu/promals3d/promals3d.php

I agree, use PROMALS3D it allows you to use protein sequences and corresponding pdb files.

Edwin

or yet " MAFFT version 7 Multiple alignment program for amino acid or nucleotide sequences"

- http://mafft.cbrc.jp/alignment/server/

* This algorithm is excellent

Hi Pedro,

for what I understand of your problem, it may be worth trying with M4T, which is probably better than MMM approach with any combined method for alignments (both servers are in Fiser's lab):

http://manaslu.aecom.yu.edu/M4T/

The problem is that you probably wish to use your own templates, so you can contact Andras.

If residues are not conserved or not fully conserved, including cysteine, you might need to consider modifying matrix substitution. For instance, substitution of cysteine are usually not favorable however from my experience i observed substitution of cysteine bridges by pairs of other residues (small hydrophobic/large hydrophobic, like G/W, G/H or A/Y). Also when binding ions, they can also be substituted by His, Asp and less frequently Glu. So aligners using classical substitution matrix (no matter the aligner) will fail to recognize optional variations specially in the case of cysteine residues.

Pedro,

Unless I misunderstood, you may find the tool called MPSA2MDSA useful. Here's a link to the paper: http://bioinformatics.oxfordjournals.org/content/23/19/2648.full.pdf

Hi Pedro,

First of all the sequence alignment (even structure-based) will be meaningful ONLY if there is some degree of sequence conservation above random, otherwise you will have a representation of a structural alignment where two residues have the same place but that won't tell you a thing (unless only very few residues are conserved eg catalytic but you'd know it) -ie in such a case some structural parts may be superposable and these parts of the proteins may have a common descent but the precise residues superposed don't have a common evolutionary, structural or functional origin- too much has changed.

So it is CRUCIAL that you use a software to check whether there is significant sequence conservation, and in what part it occurs. This is a point little understood and little put in practice but I can't stress enough its importance. The only software I know for that is HHalign http://toolkit.tuebingen.mpg.de/hhalign from Johannes Soeding's group. It is part of a great suite of tools, the MPI toolkit (HHpred, HHblits, csblast... try them). So in the top window of HHalign you put an alignment of your sequence and of sure homologs (eg found by psiblast), and in the down window an alignment of the other sequences you try to compare them with. Then you press submit, and HHalign will tell you which part of the sequences have some conservation above random, if any, and an E-value (it will convert the two alignments in profiles between the scenes and compare them). It scores secondary structure conservation too and not only sequence conservation, so if it returns no similarity, what you're trying to do is meaningless/impossible. A trick: if your query sequence is a singleton, HHalign still needs an alignment as input so paste twice your query sequence in the top window.

IF HHalign detects some similarity, then in addition to Promals3D which is an excellent suggestion (it has the great advantage of displaying the 2dary structure), I highly recommend EXPRESSO (the structure-based version of Psicoffee), which doesn't ask you the PDB numbers but does all by itself quickly. It has the great advantage of displaying the reliable and non reliable regions of the alignment (red and green/blue respectively), by comparing the results of several 3D based aligners. It is also to my knowledge the "best" 3D aligner on benchmarks, certainly made by a v reputed team (C Notredame)

Hope that helps, let me know if this works...