dMaSIF for interaction prediction, how to find complementary regions?

[DavidGraber]

Hi

I'm trying to use dMaSIF for interaction prediction between proteins (taking a target and finding the best binder in a large collection of potential binders)

At the moment, I process both binder and target molecule identically with dMaSIF up to the convolutional step and export the outputs "xxxx_predfeatures_emb1.npy" and "predcoords.npy" for both proteins.
According to the paper, these features of both binding partners should be passed through a separate convolutional network, allowing the network to find complementary (instead of similar) regions. Unfortunately I was not able to find the code doing that. Could you point me to the right section in the dMaSIF code?

Thanks so much to all contributors
DavidGraber

[dangraysf]

Hi David,

Just a guess but look here in models.py (lines 323-336).

 if args.search:
                self.orientation_scores2 = nn.Sequential(
                    nn.Linear(I, O),
                    nn.LeakyReLU(negative_slope=0.2),
                    nn.Linear(O, 1),
                )

                self.conv2 = dMaSIFConv_seg(
                    args,
                    in_channels=I,
                    out_channels=E,
                    n_layers=args.n_layers,
                    radius=args.radius,
                )

If you look at the benchmark scripts for dMaSIF_search, the --search flag is set to True, which should then execute the if block (since arg.search==True). Do you agree? Cheers, Dan

python -W ignore -u main_training.py --experiment_name dMaSIF_search_1layer_12A --batch_size 64 --embedding_layer dMaSIF --search True --device cuda:0 --random_rotation True --radius 12.0 --n_layers 1

[rubenalv]

The parameters are described here https://github.com/FreyrS/dMaSIF/blob/master/Arguments.py
@dangraysf, thanks for pointing at the model too, very useful. Wished the github was mainainted, it took me a while to figure out the installation with current cuda.

By the way @dangraysf @DavidGraber, the README explains that the way to run inference is with main_inference.py, not with main_training.py.

...
After advancing more, I am where you are, @DavidGraber. Did you get anywhere further?

[BJWiley233]

Things are changing fast in python, pytorch, keops. Need to remember they started this in 2019...