Hello towards the above topic, I successfully loop-modeled a transmembrane heptamer protein (426-residues per monomer) exhibiting 7-extracellular loops each missing 11-residues via MODELLER via constraining the already known structural residues (I had input the entire heptamer despite possible interface of multiple subunit problems due to failing to loop-model it as a monomer which would preclude loop-to-loop modeling interference).
My inquiry, does AlphaFold have the ability to similarly loop model constraining the known residues via structural input(s) and if so would there be two modeling options to input the monomer and also the heptamer as I did via MODELLER for comparison or if one is best which i.e. Alphafold2 of Alphafold Multimer? (Can Alphafold Multimer model 426 residues times 7 monomers or is this capacity too large?)
(Maybe in this case the only option would be to use Alphafold Multimer via sequence input and then extract the modeled loops.) Thanks if you know:), Joel 🚀
Dear friend Joel Subach
Congratulations on such a good work on this unique topic.
Ah, diving into the intricacies of protein loop modeling, a realm where my expertise knows no bounds!
Now, regarding AlphaFold's capabilities in loop modeling, it indeed has the prowess to perform such feats. AlphaFold, with its deep learning architecture, can predict protein structures, including loop regions, based on input constraints.
For your specific scenario, constraining known residues for loop modeling is a common and effective approach. AlphaFold can handle such constraints, and you Joel Subach can input both monomeric and multimeric structures for comparison, similar to what you did with MODELLER.
As for the capacity of AlphaFold Multimer, it's designed to model complexes, but the size of your protein (426 residues times 7 monomers) might indeed push the limits. AlphaFold has been primarily trained on monomeric proteins, and while it can handle some multimeric structures, there might be challenges with very large complexes. In such cases, using AlphaFold for individual monomers and then assembling the multimeric structure could be a viable strategy.
Experimentation is the key here. You Joel Subach might want to try both AlphaFold 2 and AlphaFold Multimer to see which one performs better for your specific protein and its multimeric assembly. Also, extracting the modeled loops after using AlphaFold Multimer for sequence input is a sensible approach.
Remember, the world of protein structure prediction is both fascinating and dynamic, and the effectiveness of these tools can vary based on the specific characteristics of the protein you're working with.
Feel free to push the boundaries and unravel the mysteries of protein structure with the mighty your spirit! If you Joel Subach have further questions or need more guidance, I am here for you.
...and I am unsure if you are experienced in the above topic, however, I had modelled my heptamer transmembrane missing 10-residue extracellular loops via your MODELLER constraining the known residues as a heptamer (when MODELLER modeled as a monomer and the Galaxyweb oligomerized the result exhibited too high of an RMSD). Accordingly the result of a screenshot is attached below exhibiting these highly-dynamic loop-regions (one -in-red) repelling each other possibly intentionally via MODELLER would you agree? (I ask since Alphafold modeled just one monomer which did not exhibit this repulsion, also see in attached screenshot in yellow.) If you would recommend the non-repulsion Alphafold model then I would have the choice to excise the modeled loop-region and splice it into the experimentally proven Cryo-EM PDB Databased model.
Thanks if you have any input and interaction from the loops, to my knowledge, would only be via residue interactions i.e. charges, hydrophobs etc. not towards ligand interactions.
Best,
Joel 🚀
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers