Please help me, the protein PDB file I predicted with D-I-TSER does not dock with the ligand

[lxp2695901]

Hello, I used the protein PDB file predicted by D-I-TASSER and put it in the autodock software. I wanted to dock with the ligand (compound), but when I followed the procedure to put the pdbqt file in, it appeared that the ligand and the receptor were too far apart, and the grid box could not wrap the ligand and the receptor at the same time. Although in the end vina was able to dock the ligand and the receptor, but put the docking file in pymol, they do not have binding bonds, so I think it should be that the protein file predicted by D-I-TASSER is different from the previous I-TASSER. Because the protein I predicted with I-TASSER before can be docked with the same procedure. Does anyone know the specific reason?Can you help me solve this problem?Thank you very much.

[ITASSERteam]

This is an interesting observation. However, docking performance can be influenced by various factors, including the quality of the receptor structure model and the specific docking algorithm used (in your case, AutoDock Vina).

Both I-TASSER and D-I-TASSER are designed primarily for monomer structure prediction, and neither explicitly considers ligand binding sites or pocket formation during modeling. In your case, the likely reason why I-TASSER gives better docking results is that it relies more heavily on template-based modeling. These templates often preserve the structural features of known binding pockets, which can facilitate docking. On the other hand, D-I-TASSER relies more on deep learning to generate structure models, and in many cases, it does not incorporate detailed homologous template information, especially for local binding sites.

Therefore, the observed difference in docking outcomes is expected and not necessarily due to a flaw in D-I-TASSER, but rather a reflection of the modeling strategies they use.

[lxp2695901]

So how should I solve the docking problem? Because I am now using autodock software to dock the proteins generated by D-I-TASSER, and there is a problem that the ligand and protein are too far apart. Should I change the docking software? Or should I modify the PDB file generated by D-I-TASSER using other methods? Do you have any good recommendations?

[XiZhang]

Thank you for your message.

The issue you're encountering, where the ligand and protein are too far apart, is common when using predicted structures from tools like D-I-TASSER. Before docking, it’s usually a good idea to first manually move the ligand close to the binding region of the receptor.

You can use PyMOL to do this: simply load both structures, and either manually drag the ligand near the protein or use a simple script to translate the ligand closer. This often helps the docking software focus on the correct binding site.

In addition, you might consider trying other docking programs like AutoDock Vina or Edock, which can sometimes offer better flexibility or sampling compared to AutoDock, especially when dealing with predicted structures.

Let me know if you need help with any of these steps!

[lxp2695901]

Thank you very much for your reply. The reason I use autodock is to calculate the parameters of the grid and prepare for the autodock vina docking. Because vina is very stable and accurate, I want to continue to use it, but now the autodock docking cannot be completed. When the grid box step is reached, the ligand and protein will appear. The distance between them is too far, and the docking pocket cannot cover the distance between them. Therefore, I cannot calculate the parameters of the grid, so I cannot perform the docking of vina, which makes me very distressed.Do you have any good solutions to this?Thank you very much for helping me