bond angles, dihedral torsions and atom-atom non covalent contacts.

2. Check the folding pattern in the search for knots of combinations not found in the PDB database. You can use servers such a DALI to find partners for your pattern, if you can find anything, be suspicious!

3. Find if your model is Biologically appropriate. I.e. if the 3-D coordinates of the backbone atoms are fit to host the side chains of the amino acid sequence your intended to model.

For the first issue you may use MODELLER, as recommended by Sandipan Chowdhury above, or any other physics-based forcefield potential, such as gromos, which is conveniently packed into swisspdbviewer, and is easy to use, because it takes care of informatics details such as topology assignments, which can be quite a nuisance. Be sure to look at the detalied by-residue energy report, and spot anomalous high energy sites. You can also look at the what-if suit of programs to search for anomalies and the many other tools proposed above by Sinosh Skariyachan.

For the non-covalent interaction of neighbours a very good choice is ANOLEA energy (as commented by Arpita Batta above), also available at the swiss-model portal (mentioned by Sandipan Chowdhury above) where additional tools are included.

For the second part, there are a number of structural alignment programs such as STAMP (see VMD site: http://www.ks.uiuc.edu/Research/vmd/), or TOPOFIT (see the above recomended paper for a proper citation.

For the 3rd issue, the best option (unique because of the information it provides) is the Rd.HMM protocol (read: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0012483). If you need help with this, you may contact me directly by e-mail.

Additional information can be found at:

http://www.cicy.mx/Sitios/Journal/pdf/2011/October/HighDefinition/IBCJ_2011_1_1_26S.pdf

A reading i truly recommend for novice modellers.

(1) Modeller has a knowledge-based structure scoring function called "DOPE" which can be used to score multiple homology models generated.

(2) PROCHECK is a program which assess the stereochemical quality of the structure based on several structural constraints on protein structures (for e.g. how many bond angles fall within allowed regions of the Ramachandran plot - fewer the bond angles within disallowed regions better the structure)

(3) An assortment of such computational tools can be found here:

http://swissmodel.expasy.org/workspace/index.php?func=tools_structureassessment1

(4) One other way, although not as straightforward as the others listed, would be to use MD simulations. A good model, upon running a (simple) molecular dynamics simulation, will quickly (a few nanoseconds at max) reach a saturating RMSD or energy value. Poor structures will either never saturate or will have very high RMSD/RMSF values or will take much much longer to reach an equilibration.

Unfortunately, however useful these above techniques are they can never be used to argue the "correctness" of a model. It can be used to compare one model vs another but that's about it.

Hope this helps. Good luck :)

helo sir you can use the anolea server for this. This link might be of some help http://protein.bio.puc.cl/cardex/servers/

and

www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CGAQFjAA&url=http://protein.bio.puc.cl/cardex/servers/anolea/&ei=BEUpUMX5C4_jrAe2ooCICg&usg=AFQjCNHr3H0-tD-hdCX5M0KfNw46PXMOAQ&sig2=8eEsfF3MNYj1jk06IGfvBw

Acelerys.... has free time period before requiring purchase of licence

verify_3d server will give you procheck, errat ramachandranplot..etc. so this server is better to find modeled structure accuracy

There are some computational tools which works very well to find out the quality of the model. The most popular programmes are PROCHECK, WHAT CHECK, VADAR, DSSP, ERRAT, GROMOS, CHARMM, ANOLEA VERIFY3D, PROSA, PROVE, YASARA etc. MODELLER Software also has many tools to find out quality of the modeled structure!

bond angles, dihedral torsions and atom-atom non covalent contacts.

2. Check the folding pattern in the search for knots of combinations not found in the PDB database. You can use servers such a DALI to find partners for your pattern, if you can find anything, be suspicious!

3. Find if your model is Biologically appropriate. I.e. if the 3-D coordinates of the backbone atoms are fit to host the side chains of the amino acid sequence your intended to model.

For the first issue you may use MODELLER, as recommended by Sandipan Chowdhury above, or any other physics-based forcefield potential, such as gromos, which is conveniently packed into swisspdbviewer, and is easy to use, because it takes care of informatics details such as topology assignments, which can be quite a nuisance. Be sure to look at the detalied by-residue energy report, and spot anomalous high energy sites. You can also look at the what-if suit of programs to search for anomalies and the many other tools proposed above by Sinosh Skariyachan.

For the non-covalent interaction of neighbours a very good choice is ANOLEA energy (as commented by Arpita Batta above), also available at the swiss-model portal (mentioned by Sandipan Chowdhury above) where additional tools are included.

For the second part, there are a number of structural alignment programs such as STAMP (see VMD site: http://www.ks.uiuc.edu/Research/vmd/), or TOPOFIT (see the above recomended paper for a proper citation.

For the 3rd issue, the best option (unique because of the information it provides) is the Rd.HMM protocol (read: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0012483). If you need help with this, you may contact me directly by e-mail.

Additional information can be found at:

http://www.cicy.mx/Sitios/Journal/pdf/2011/October/HighDefinition/IBCJ_2011_1_1_26S.pdf

A reading i truly recommend for novice modellers.

Please add this at the beginning of the above comment. I seem to have deleted accidentally:

this is a multifaceted problem.

1. check a structural details to conform to canonical know values, i.e.

Very helpful and interesting post. To asses the quality of molecular models you can also use MaxSub ( http://www.cs.bgu.ac.il/~dfischer/MaxSub/)

Ultimate method:

-make a cristal

-use difraction of X ray

-compare

then... no need of any model

Never forget that accurency of a protein model is depending on the sequence: If you use a sequence that has already many homologs in database, it would be accuarate, if you do complete new sequence.... it will be random...

you can use SAVES server for structure validation.

OR

you can used PROSA server.

There are a set of programs called, model quality assessment programs MQAPs, which have become popular as a result of the CASP competition. MQAPs produce a global score between 0 and 1 in relation to model quality, with several methods such as ModFOLD and QMEAN also producing per-residue errors.

You might want to take a look at the "quality estimation" tool of the Protein Model Portal (PMP). It includes/compares various evaluation tools such as ModEval, ModFold, QMEAN.

Try using PROCHECK - a 3D structural analysis program written by I believe Janet Thornton et al.

This is the web-site :

http://www.ebi.ac.uk/thornton-srv/software/PROCHECK/

suggesting to use web based 3D structure validation at http://nihserver.mbi.ucla.edu/SAVES/

If you have an idea of which family this belongs to you can check out evolutionary relationship using msTALI

A very effective method to check the quality of protein model may be used following the method described in my paper below. If needed, I can help you to estimate the quality of your protein structures / models.

http://onlinelibrary.wiley.com/doi/10.1002/pro.2007/abstract

After checking the accuracy of your modelled structure, you should carefully compare your structure with experimental data. Bioinformatic tools may help you detecting problems within your structure as side chains clashes or hydrogen bonds being not formed and of course also in scoring the quality of the model. However, you will probably not get a final clue whether you are close to the native structure, until you compare your models with experimental data i.e. point mutations. If such data are available they provide very strong arguments to select one model against another.

Janet Thorntons group published a very useful protein structural assessment tool called Procheck some years back. Please have a look at this web page:

http://swissmodel.expasy.org/workspace/tutorial/geometry.html

Another simple but sensitive way to filter bad modeled structures from god ones, is to screen for buried charged residues. If you find them - you have to check the structure carefully - it is very expensive energetically to bury a charged residue. If it is found to be involved in a salt bridge, it is most likely OK. If not discard the structure .

Procheck is the first step you must do. You access it online from various web servers or you can also use PDBsum, it will show procheck results and some other result also.

AFM could be applied for some visualizations, for more detail read attached publication:

Article AFM study of complement system assembly initiated by antigen...