Recommended receptor/GPF preparation workflow for AutoDock-GPU and difference in dielectric values generated by Meeko vs ADT

[Degax]

I am preparing receptor and grid files for AutoDock-GPU docking, and I would like to ask for your recommendation regarding the proper receptor/GPF preparation workflow.
According to the AutoDock-GPU documentation, AutoDock-GPU requires Meeko for preparing receptors and ligands, and AutoGrid for calculating the affinity grid maps, including the .maps.fld file passed to --ffile.
In my case, I am docking ligands to CYP3A4 WT and CYP3A4 variants. Since I need to compare docking results between WT and mutant proteins, I want to make sure that the receptor preparation and grid generation steps are consistent and appropriate.

I noticed one difference between GPF files generated by different tools:

When I prepare the receptor and GPF using Meeko / mk_prepare_receptor.py, the generated GPF contains:

dielectric -42.000

However, when I prepare the receptor and GPF using ADT / AutoDockTools, the generated GPF contains:

dielectric -0.1465

From the AutoDock/AutoGrid documentation, I understand that a negative dielectric value indicates the use of the Mehler and Solmajer distance-dependent dielectric function, whereas a positive value would be interpreted as a constant dielectric. I also found in the AutoDock Vina/AutoGrid tutorials that when the dielectric value is negative, AutoGrid uses the distance-dependent dielectric function regardless of the exact negative number.
Therefore, I would like to ask:
For AutoDock-GPU, is the recommended workflow to prepare the receptor and GPF using Meeko / mk_prepare_receptor.py, rather than ADT / AutoDockTools?
In the AutoGrid4 + AutoDock-GPU workflow, are dielectric -42.000 and dielectric -0.1465 functionally equivalent because both are negative and therefore trigger the distance-dependent dielectric function?
Should I keep the dielectric -42.000 value generated by Meeko, or should I manually change it to dielectric -0.1465 for consistency with classic AutoDockTools-generated GPF files?
For comparative docking between CYP3A4 WT and CYP3A4 variants, would it be acceptable as long as all systems are prepared using the same pipeline, same ligand preparation method, same grid box center/size/spacing, same atom types, and same AutoGrid4/AutoDock-GPU parameters?

[diogomart]

For AutoDock-GPU, is the recommended workflow to prepare the receptor and GPF using Meeko / mk_prepare_receptor.py, rather than ADT / AutoDockTools?

Yes

-42.000 and dielectric -0.1465 functionally equivalent because both are negative and therefore trigger the distance-dependent dielectric function?

Yes

Should I keep the dielectric -42.000 value generated by Meeko, or should I manually change it to dielectric -0.1465 for consistency with classic AutoDockTools-generated GPF files?

It doesn't matter. You can choose your favorite negative number. Ours is -42 ;-)

For comparative docking between CYP3A4 WT and CYP3A4 variants, would it be acceptable as long as all systems are prepared using the same pipeline, same ligand preparation method, same grid box center/size/spacing, same atom types, and same AutoGrid4/AutoDock-GPU parameters?

What is it that you are asking the acceptability of?

[Degax]

Thank you, sir. You have already helped me solve most of my questions.
I have one more question. When I prepare the receptor and ligand using ADT and then perform docking with AutoDock, the binding energies I obtain are normal negative values. However, when I prepare the receptor and ligand using Meeko and then perform docking with AutoDock-GPU, the binding energies become positive values.
Could this be caused by differences between Meeko and ADT in receptor/ligand preparation, or is it more likely due to an issue in my own operation or setup?

[diogomart]

I would have to look at the files to be sure, but my first guess is that the search space (a.k.a box or grid) are different. You could the check the header (first 6 lines) of map files written by autogrid to confirm that the center and number of grid points are the same from ADT and meeko.