What is the difference between molgrid in libmolgrid and gnina? about libmolgrid

[dkoes]

libmolgird has factored our gridding code into a standalone library. The current version of gnina uses libmolgrid as a dependency.

We need to do a little bit more work with documentation and adding a few features before the first release, but you can use libmolgrid as it is now. See:
https://github.com/gnina/libmolgrid/blob/master/test/test_torch_cnn.py
for a pytorch example.

[yamasakih]

Thanks so much for your prompt help!

I'm very glad to have an example because I thought that I implemented it with Pytorch.

In addition to libmolgrid, should I compile gninagrid and preprocess the data?

[dkoes]

You do not pregenerate grids - that's the point of molgrid to generate them on the fly. It's important to be able to have random rotations and translations. You can preprocess molecular data into gninatypes files and a create a single molcache with create_caches2.py for faster training if you are fine using gninatypes.

[yamasakih]

Thank you for the detailed explanation.

I only have to use molgrid. It's very easy :)

There are many useful functions in the gnina/scripts repository.
I will look into it for study.

Thank you very much!