The best way to build a model depends on the degree of homology to the best templates available in the pdb. If a blast search of the PDB (https://www.rcsb.org, advanced search mode) finds reasonably close homologs, Swiss model (https://swissmodel.expasy.org) is a good starting point. For poorer fits, a threading algorithm like iTASSER (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) may give you better results. For low homology templates, other methods exist that are not fully automated, but require more input from the user. Given that membrane proteins provide a special challenge, it is a good idea to try different methods and compare the results.

Article Homology modeling of membrane proteins: A critical assessment

The best platforms as far as I know are

Modeller

Robetta

i-TASSER

Swissmodel

If you have you want to model a loop region of a protein, you can use Modeller. If this region contains an alpha helx or a beta sheet (prediction by PSIPRED or any other platform), you can specify that while using Modeller. Modeller can also perform threading. For example, you want to model X, which is homologous to Y and there is a structure of Y available on PDB. In that case, you can use the available structure of Y and integrate the sequence of X.

If you want to model (predict) a completely new protein which does not seem to have full-length homologs, then you may use Robetta and iTASSER. I suggest you use both, gain as much knowledge from literature to validate the structures, and decide which one's better. You can also compare the predicted structures with their secondary structure predictions by PSIPRED and see which structure matches best. In my experience, Robetta turned out better. It's always good to have as many predictions, hence Swissmodel is suggested too because it is another good standard for making predictions. At the end of the day, I would also suggest that you simulate your protein model(s) and see how stable they are.

Hope that helps.

You can predict the 3D structure (PDB format) of the desired protein by utilizing Modeller, Swiss Model, or Phyre2.

Abu Saim Mohammad Saikat, Dheeraj Prakaash AND Annemarie Honegger thank you so much!

http://journal.umpalangkaraya.ac.id/index.php/jmd/announcement/view/17

https://docs.google.com/forms/d/e/1FAIpQLSfNpiah9wUc--jA7V1_42J2M3Cjlgtvgj-vDEasBIPXbEs9MQ/viewform?usp=pp_url

Dear Colleagues,

Department of Pharmacy, Universitas Muhammadiyah Palangkaraya openly inviting researchers with expertise in the field of Molecular Docking to join the new journal that will be released in early 2021, both as an editorial board and as peer-reviewers.

We will call this journal "Journal of Molecular Docking" or "J. Mol. Docking", with a very specific aim and scope, namely various research in the field of science and health using Molecular Docking as its main method.

Currently we are still preparing the journaling platform with OJS 3.1.1.2 and will start collecting articles as soon as the editorial board and peer-reviewer members are gathered.

It should be noted that this position is voluntary, so all editors and reviewers will not receive any rewards in the form of funds. However, we will provide a certificate of appreciation for each manuscript handled, both as an editor and reviewer.

The target of this new journal is to be indexable by DOAJ after the first year and indexed by Scopus and/or ESCI WoS after the fifth year.

For this reason, we expect the willingness and participation of Molecular Docking researchers to be directly involved in the development of this new journal so that later it can become a publication benchmark for research with the molecular docking method.

Once again, we are grateful for the participation of all researchers and hope for their support to be able to create and grow this new journal.

Best regards,

Mohammad Rizki Fadhil Pratama

Department of Pharmacy

Universitas Muhammadiyah Palangkaraya

http://orcid.org/0000-0002-0727-4392

Thank you so much Aniket Kumar

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!!

Anuraj Nayarisseri thank you very much for your suggestion.

One of the reasons why I suggest SwissModel for a first try is because if you do the "search for templates" first, it not only gives you a list of potential templates, but annotates these to their oligomeric state, presence of ligands, link to the UniProt record etc. and allows you to view a 3D alignment of these templates to see whether these have an influence on the conformation. It the template is multimeric and the interface seems to be conserved, the models generated by SwissModel on an oligomeric template respect this and free-model missing loops in such a way that they do not clash, while for example with iTasser the same task using the same template yielded mostly models in which the loop clashed upon oligomerization, and hid an epitope that from lab results I knew to be accessible.

Depending on the purpose of your model, it can be very important to take into account potential conformational changes due to ligands and/or oligomerization.

iTasser and other threading-based algorithms are better for cases where already the sequence alignment is uncertain - in this case you anyhow cannot trust on ligand binding sites or protein/protein interaction sites being conserved.

Alphafold can sometimes do funny things with natively unfolded stretches, e.g. in a GPCR stick the N-terminus of the receptor into the ligand binding site and the C-terminus into the pocket which would usually accommodate the C-terminus of G-Alpha

Annemarie Honegger Thank you very much for your valuable comments.

Annemarie Honegger has amazing comments! Also, if your protein is less than 200AA, you can use QUARK - as it is a computer algorithm for ab initio protein structure prediction and protein-peptide folding so no templates are needed.

Andi S. Silva Thank you very much.