you can use several templates, even one for each part of the protein. I recommend you EasyModeller, a GUI for modeller that simplifies the work with Modeller. You can use several templates, and previously remove the non-coinciding part of the pdb file, in this way, you use appropiate template for each part in the alignement

good luck

I had the same problem some years ago. I wish I had known then. Thanks.

The effect of residues at the N-terminus of a protein is structure dependent. In many proteins the N-terminus is flexible and has very little impact on the overall structure. However, in some cases, the N-terminus makes important interactions that are critical for proper folding of the structure.

For examining the effect of different amino acids at the N-terminus a structure prediction program can be very helpful. I highly recommend iTASSER ( http://zhanglab.ccmb.med.umich.edu/I-TASSER/ ). At a minimum, all you need for a structure prediction with iTASSER is the amino acid sequence (in this mode iTASSER will automatically select templates). If there is already a structure of your protein of interest (or a homolog) then you can instruct iTASSER to use that structure as a starting point for the modeling.

For your scenario, try modeling runs with iTASSER for each of the different sequences.

In most cases the peptide signal is absent from the template pdb file (often removed for crystalisation, sometimes not ordered in the crystal). There is therefore a chance it does not make a difference...

In any case, one option is to prepare the model only for the part of the protein for which you are confident, i.e. without the peptide signal. That avoids overinterpreting a wrong model...

Thanks everyone for your valid suggestions. But the protein sequence similarity with existing protein structure is around 98% is it advisable to opt for threading modelling using i-TASSER than homology modelling?

with 98% similarity you should go for homology modeling. But you might still need to construct the peptide signal if, at that place, the similarity is low.

If you could elaborate on your question it would be more helpful. However I have always felt that a threading modelling approach like i-TASSER is always better than any homology modelling approach even when the similarity is 98% because even with a threading approach the homology of the sequence is often taken care of. Also I really think you should predict three different models and strcuturally align them to see if they are really different from each other or not and if yes at what locations. This will tell you if the additional aa influence the structure or not. Also, before you proceed with modelling a bit of background check(for all three sequences) in terms of domain prediction and possible substructure motifs in your sequence which can be done by a host of prediction servers online(like SEGMER,Domaination) will also give you more information on the direction you should proceed towards.

The extra a.a residue at your N-terminal will make a difference to your protein structure . And actually this can be a good strategy to go ahead with combination of thread modelling and homology modelling.

For thread modelling you can use i-TASSER or Phyre2 Server, they both generate quality results.

Best of luck!

You can use homology modelling using SwissModel, or Rosetta. Try the different sequences (i do not understand why you have so many sequences, isoforms?).

And as model usually you should use the one that has the highest sequence homology with the protein with known structure.

with the different sequences try to predict the secondary structure to see if they differ, probably not. The N-terminal sequences are the most different regions in a protein, but may not influence its ternary structure.

Antoine Taly Thanks for your advice. But the extra aa residues are also the part of peptide sequence right. So what is the possibility that neglecting these residues will give appropriate structure and not a incomplete structure?

Thanks everyone I will opt for easymodeller

Modeller(Sali Lab) has an advanced homology modelling option whereby you can use an alignment of multiple templates to predict a structure. You can try that.

The Protein Model Portal's Interactive Modeling feature will find the best template and and package the request to send to 6 modeling servers. many of which were mentioned above. All results are emailed to you, and you can compare and contrast all of the results. You can use this Portal to see if any of the participating modeling groups have already created one anyway too by searching by sequence.

http://www.proteinmodelportal.org/?pid=modelling_interactive

The impact of extra amino acid at N-terminal depends on the type of protein.. Say for eg: if the n terminus of the protein is involved in acetylation this definitely would affect the protein-protein interaction inturn changes the protein function and thus inturn the protein will have a different structure based on the STRUCTURE-FUNCTION relationship.

I-tasser is a good start for modelling but a online server with less or no parameters might not give you a precise structure. So I would recommend using " MODELLER " http://salilab.org/modeller/ or Schrödinger's Maestro module (if you have access to) which is an excellent software for modelling. In both softwares you can search for template structures for all 3 AA sequence you have. My guess is you might end up getting the same template for all three if the protein is not highly characterized.

Good luck.

Hi,

There is no right way of doing this. Mostly, and by this i mean only the structures i have seen or have worked with, the terminal residues form loops (disordered structure). I think from what i have learned so far is that this should always be the case since its is very easy to modify disordered structures rather then ordered. Modification is required I think when the protein has to be degraded. I am not a biologist or a chemist or a bio chemist so i might be wrong but this is true form what i have seen. Now, the core of the protein is what folds and shapes up and presents a binding site. In instances where N terminal residues are binding they again are disordered.

So after the story above - i dont think that the N terminal residues should effect the structure of the core or for that matter play any part in the prediction of the structure which you will be predicting using a prediction software.

So bottom line - use any ... to test the hypothesis - generate 3 structures compare them with each other. Notice the core would be more or less identical.

Also note that the above applies only if the N terminal of the three sequences you have is the N terminal of the protein and not just an intermediate residue which you think is the N terminal ...

Good Luck ...

The best solution to this problem I found is in this paper:

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0012483

First use sequence similarity to spot possible templates in PDB.

Use blast to seek for "highly similar ones", use psi-blast for "somehow similar cases" and Delta-Blast for "distantly related".

Yo can also use the HMMsearch tools to perform your search. The easiest way is to use the HHpred tool (http://toolkit.tuebingen.mpg.de/hhpred) at Max Plank's Institute servers.

Next take the PDB files selected as "possible tempates" use rosetta design (https://www.rosettacommons.org/software/) to generate a 100-150 sequences using a Res file with allaa flag, for all positions.

This means, rebuild the sequence using any possible residue at every position in the protein structure. DO THIS WITH THE FIXED-BACKBONE OPTION if you allow for flexible backbone you will end up with garbage.

Extract all amino acid sequences from the Rosetta-design artificial PDB's

put them in fasta format in a plain text file.

Nest you need HMMer tools (http://hmmer.janelia.org/), use HMMbuild to generate and calibrate a HHMer model (*.hmm) then use HMMsearch to defy the model against a protein sequence database (a fasta-formated protein sequence file with the amino acid sequence of the file you intend to model, usually the amino acid sequence of the protein which was solved and stored in the starting PDB file, and as many more other seqeunces as you want, may be the whole refseq, nr or UniprotSprot, if you care to download them to your disk).

If the PDB template, you chose to begin with, is a good as template, then HMMsearch will include it in its output and will give it a large positive score and low expectancy value (E-value). In the file you will most likely find the amino acid sequence of the PDB putative template in the first top 5 hits. A good score should be as large as half the length of the amino acid sequence, or above. In my experience, you only get this for the amino acid sequence of the X-ray solved protein .

THIS METHOD DOES NOT GIVE FALSE POSITIVES, BUT CAN GIVE YOU FALSE NEGATIVES.

In other words if it gives you a positive answer (scores your sequence with a positive score and very low Expectancy vales). Then you can be sure of the template adequacy. A negative answer does not means the template is useless, it just mean you do not know for certain how good it can be.

Once modeled, you can use the same strategy to check is the model is better for your sequence than for the template. In many cases is the other way around, meaning, you simple copied the template folding without fitting the structure to accommodate the new amino acid sequence.

I strongly suggest you read the paper.

best wishes,

Rogelio

From your description I understand that you're working with sequences very similar to each other and so most likely they will show a very high sequence identity. If so, they present a similar tertiary structure (see principle homology modeling).

You can use several strategies to create your final model, this depends on what you're looking for.

1) of course you can create a template for each sequence and evaluate the energy stability (Gibbs) of each of them.

2) Or you can create a template for each sequence, and through a multiple alignment order to create a final model.

Regarding the influence of the final residues in the N terminal on the stability of the model depends on their number and their arrangement along the chain. It is a membrane receptor?. though few are unlikely to destabilize the structure of the protein. A secondary structure such as alpha helix needs at least 6 aa. You also have to understand what kind of question is Cys residue? could act as a bridge etc..

Mine are only general guidelines to get a detailed answer must be analyzed sequences under consideration.

Good job....

Try homology modeling using SWISS MODEL.

You can check the available protein sequence from Uniprot database of your protein of interest. From this complete aa sequence of your protein of interest might be identified.

Functional protein after it reaches its target location, lost its signal sequence from its N-terminal end. So, your protein might have signal sequence as additional protein from your protein. You can identify the signal sequence by using signal sequence prediction servers like SignalP, PrediSi. After identifing signal sequence, they might be removed and utilized the remaining sequence for prediction.

Template for your protein might be determined using SWISS MODEL optimized option.

Hi there, try our server (link below) and submit each of your sequence sets.

http://www.reading.ac.uk/bioinf/IntFOLD/IntFOLD2_form.html

HTH

Pl take consensus of the protein sequence and try in i-TASSER or Swiss Model

Hi,

To select a template which is appropriate for protein structure development is easy. First of all, I hope you know about a tool named- Basic Local Alignment Search Tool (BLAST).

Try out BLASTp/ SMART BLAST so that you get a list of proteins which have similar sequence identity having a good E-value.

Once you done with it, you can click on the best option retrieved and you will be directed to the templates main page.

Or,

You can use SWISS-MODEL tool, as it has two options, i) Template development & ii) Model development.

You can click on the template generation button and get your job done.

Best regards,