Modeller is a homology modelling tool and it should be able to do that as well. Here is the web page: http://salilab.org/modeller/

http://salilab.org/modeller/

Thanks Serdar. I tried with Modeller before. However, the accuracy of loop prediction by Modeller is low compared to other loop prediction softwares, as per benchmarking.

Hence, I was searching for something better than modeller. 

Hi Abdul, thanks a lot for this new piece of information. I have downloaded DisGro. About to start the modeling. Once I get the results will inform.

Are you building into a map? Or, are you taking a published structure and adding loops for some future step?

Either way, you can download the experimental map, along with the PDB, and drop the sigma to "see" where the loops are most likely going (usually build at 1 sigma for 2Fo-Fc and build into spotty density at 0.7 or so). In coot, I would usually perform a real space refinement including the recently built loop and anchors where the density is well resolved (say two residues preceding the unmodeled loop and two residues after the unmodeled loop) after modeling some residues into the 0.7 sigma map. I have built numerous loops in this manner and received little criticism since I used the experimental map for my modeling. Using the fit_loop module in Phenix you should also be able to get to some fairly reasonable models.

Here is a link to numerous loop-modeling software packages. Some are web-server based and some require local installation.

http://www.vls3d.com/index.php/links/bioinformatics/3d-structure-prediction/modeling-loops

Again, I believe that it is critical to model loops using some experimental data to increase the quality of the model and reduce the criticisms that arise from protein modeling in general. I used the aforementioned building technique (using the experimental map) to model loops that were subsequently subjected molecular dynamics in GROMACS with success. A couple rounds of energy minimization would be beneficial upon completion of modeling the loops that are either derived from your building into density or homology modeling servers to reduce unnatural loop conformations (GROMACS is easy).

Good luck.  

Dear Carter,

Thanks a lot.

My problem is neither building the loop into a map nor I am adding those to a published structure.

I have model built three different pieces of coiled coil dimer. Those dimers are connected to each other by linker regions which could be treated similar to loops.

Quality of those model coiled coil structures are verified by some other means. Now I want to add those linker regions, which should be fairly accurate.

As this protein (actually a pair of proteins coexist together) exist in filamentous form, it is hard to crystallize. The protein is crystallized only in parts. The experimental data available on this protein are electron micrographs. Any idea how to exploit this experimental data?

Thanks

A buddy of mine used MODELLER for a 24 residue linker. He had to do a local install to get the results he was looking for. He also had SAXS data to help corroborate the model. Do you have any SAXS data?

The following link takes you to a 2008 paper that looks at flexible fitting of crystal structures into electron micrographs using molecular dynamics. Once convergence has met, and you feel that the crystal structures are appropriately fitted into the EM, then you can have an understanding as to where the loops HAVE to go.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2430731/

Other methods are available that do a similar sort of fitting, accept it uses Normal Mode Analysis, which is much less computationally intensive.

I hope this is helpful.

The loop modeling protocol in Rosetta has been shown to be quite accurate for rathe rlong loops. So, that might be an option, too.

http://www.rosettacommons.org

-> loop modeling documentation:

https://www.rosettacommons.org/docs/latest/Application-Documentation.html#Structure-Prediction

-> the latest paper

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0074830

Good luck

The following papers may provide clues:

A new method for the automatic determination of cores in proteins

M. Pirtskhalava, B. Vishnepolsky and G. Managadze

J. biol. Phys. Chem. 2 (2002) 1

Side chain orientations as a characteristic of native protein structure

G. Managadze, B. Vishnepolsky and M. Pirtskhalava

J. biol. Phys. Chem. 4 (2004)33

Thanks a lot,  Carter, Sebastian and Jeremy. I would try all these options. May be the comparative study will bring out the best solution