Swiss-model model quality

Question

[mbrav005]

Hey all, I have been trying to model my sequences via swiss-model and I believe if im not mistaken a good/ok model is determined by a higher QMEAN>-4.0. I am wondering how a structure can get such a low QMEAN i.e. -7.0 if the sequence identity between the template and target are 100%? For example I am taking an e.coli sequence and modeling it on a e.coli structure and the QMEAN is terrible? Am I going about this the wrong way? I am trying to get as good of a model as possible, are there other statistics to consider? The GMQE explained by swiss-model is not very intuitive but I guess it is used to rank models in your search.

Thanks

Answer 1

[Mensur Dlakic]

You should inspect the alignments before modeling and make sure that there is an actual template match between your sequence and the PDB structure. PDB files often lack portions of proteins due to unstructured termini or inherent flexibility of protein's parts. If that's the case, in those regions your model will be made without any homology restraints, and will not be meaningful. This will manifest as floppy parts that extend out of protein core like spaghetti sticking out of a meatball. If this happens to be your problem, I suggest you choose the interactive mode and to manually delete parts of the alignment lacking the template.

Out of curiosity, why are you modeling proteins with 100% identity to a known structure?

[mbrav005]

I was suspecting that it is due to the pdb lacking the full length of the alignment, I will try to adress this issue within interactive mode. Actually I have been just trying to see how swiss-model scores and models my seqeunce of 100% identity. It makes more sense in the future to just use the raw pdb without the alignment in my future modeling projects since it is most likely the same structure. Thank you for your suggestion. Do you mean interactive mode in Swiss-model?