AutoGrow4 without docking

[Wout_Van_Eynde]

Dear

Is it possible to run AutoGrow4 without docking, e.g. when the target isn’t known or hasn’t a validated structure? If yes, how would one do this and would you recommend this? I guess it is done by adjusting the .json file to that extent, but I am not too sure.

I have no experience in AI nor in programming, just a great interest in this field, so excuse me for the lack of knowledge.

Warm regards,
Wout Van Eynde, student at KU Leuven

[Jacob Durrant]

Thanks for reaching out and for your interest in AutoGrow4! The current version of the software uses a docking step to select the best molecules, but it should be possible to modify the source code to skip this step. Docking does plays a crucial role in the compound-selection process used to form subsequent generations, though, so I don't recommend skipping that step. Hope this answer helps!

[Dr. Amol]

On the same lines to the question above I take this opportunity to ask a somewhat similar question. I would like to use my own AIML based predictor instead of any of the docking programs. The AutoGrow4 research article mentions that "With the appropriate plugin code, virtually any method for docking and/or reassessing ligand binding can be integrated into AutoGrow4, including quantitative structure–activity relationship (QSAR) approaches,
ligand-similarity evaluations, etc.", however, there is not much information on how to do it. Any guidance in this direction is requested. Thanks in advance.

[Jacob Durrant]

Thanks for your interest, too, Dr. Amol. Replacing the docking programs with another metric shouldn't be too hard. Here's the class used for Vina docking, for example: https://github.com/durrantlab/autogrow4/blob/master/autogrow/docking/docking_class/docking_class_children/vina_docking.py

You might use that as a starting template to build a related class for your AIML predictor. Hope this helps!

~Jacob

[Author]

Posts