I am trying to build the model of a protein using homology modeling. The available structures of the protein are partial. Specifically, the available structures are of specific domains/segments of the protein, not the complete one. I would like to build a complete model of the protein. What would be the best approach to do so? I am trying it with Modeller but it can't properly link the domains. Is there any other specific approach to do that kind of modeling? Instead of Modeller, is there any other tool/software (preferred the free one) that can do such modeling?
Actually yes, there are servers like SWISS-model but I don't think the result of a blind prediction modeling developed in this way will be much different. Modeller's multi-template mode is very useful for this goal. You can use external restraints and secondary structure properties to improve your model.
Hello Aritra Bej
Yes, there are different online servers and standalone software which can resolve your problem, but for that, you must know the different aspects of modelling. Like you are performing homology modelling and according to your query, you are not getting any template sequence with good query coverage, sequence similarity and sequence identity. In my opinion, you should go for fold recognition based modelling where you can use different folds to create your model. Basically, you don't have to do anything the software will run by itself according to its algorithm. You just have to provide the FASTA sequence of your protein. Even after performing fold recognition based modelling and you don't find good results it is better to go for ab initio modelling, where you have to model your protein from the scratch.
There are different software which can help you:
HHpred, RAPTOR, Phyre2
I think this information might help you.
Thank you Rameez. I will definitely try my sequence with the software that you have suggested.
Hello Aritra,
Simple approach will be to use I-Tasser and Phyre, sometimes they built it.
Challenging approach: suppose your seq. length is 120 aa and you are getting suitable template of 45-120 aa and now you have to built starting 45 amino acids. So you take only this specific region of 1-45 residues and BLAST it and now if you are lucky, you may get some suitable template for this region. Take both templates and use Multiple template modeling approach in MODELLER. linking is not an issue you may have to edit the files a bit.
I would also recommend you to check disordered probability of the region that you are not getting. Use DISOPRED first to see that particular region might be disordered and you may not be able model it.
Thanks, Mitul for your suggestion. However, I have already built my model in MODELLER. The approach was nearly the same that you have suggested here.
Aritra Bej i m also facing the same issue will you please share the solution
Akhil Kumar In my case, two domains of a protein were connected by a linker. I have used Ab-initio domain assembly server (https://aida.godziklab.org/AIDA) to connect those domains using the loop.
Try to model the homologous models by any of following tools then stitch the multiple fragments using Python Programming.
Try Alphafold deep mind prediction https://deepmind.com/research/case-studies/alphafold
PS2 http://ps2.life.nctu.edu.tw/
cph http://www.cbs.dtu.dk/services/CPHmodels/
phyre2 http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index
I-TSSER: https://zhanggroup.org/I-TASSER/
Best!!
Try alpha fold from deep mind. It will predict upto 2000 residues protein.
Thank you
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python