why the interface is computed like this ?

[camel2000]

this function in data_iteration.py as below,

def generate_matchinglabels(args, P1, P2):
    if args.random_rotation:
        P1["xyz"] = torch.matmul(P1["rand_rot"].T, P1["xyz"].T).T + P1["atom_center"]
        P2["xyz"] = torch.matmul(P2["rand_rot"].T, P2["xyz"].T).T + P2["atom_center"]
    xyz1_i = LazyTensor(P1["xyz"][:, None, :].contiguous())
    xyz2_j = LazyTensor(P2["xyz"][None, :, :].contiguous())

    xyz_dists = ((xyz1_i - xyz2_j) ** 2).sum(-1).sqrt()
    xyz_dists = (1.0 - xyz_dists).step()

    p1_iface_labels = (xyz_dists.sum(1) > 1.0).float().view(-1)
    p2_iface_labels = (xyz_dists.sum(0) > 1.0).float().view(-1)

    P1["labels"] = p1_iface_labels
    P2["labels"] = p2_iface_labels

my question is: what is the meaning of xyz_dists.sum(1) > 1.0 and I think the value in the array of xyz_dists.sum(1) is less than 1.0 , and maybe in most case the value should be negative, so the p1_iface_labels may have no non-zero value. which means there is no interface.

[YAndrewL]

Same question, did you solve this?

[jeanfeydy]

Hi @camel2000, @YAndrewL ,

Apologies for the long delay. To answer briefly: please note that the .step() function is "old" KeOps syntax that means "> 0". As a consequence, writing:

xyz_dists = ((xyz1_i - xyz2_j) ** 2).sum(-1).sqrt()
xyz_dists = (1.0 - xyz_dists).step()

creates a (N_1, N_2) LazyTensor of "booleans" whose value xyz_dists[i, j] indicates if point i from protein 1 and point j from protein 2 are at distance < 1 Angstrom from each other. Then, computing:

p1_iface_labels = (xyz_dists.sum(1) > 1.0).float().view(-1)

is just a way of looking for points i in protein 1 with at least two such neighbours in protein 2. All in all, this code "flags" points on the surface of protein 1 that are close to the interface of protein 2, and vice versa.

I hope that this answers your question, please feel free to re-open the issue if needed :-) Best regards, Jean

[YAndrewL]

Hi @jeanfeydy, thank you very much for the reply. Is it fine to use the latest version of keops and pytorch geometrics?

[camel2000]

thanks very much, this problem bave been solved , Not long ago, I also checked the official website of kops, the api ".step()" is really confusing, but now it is clear 

李 @.***

 

------------------ 原始邮件 ------------------ 发件人: "Jean @.>; 发送时间: 2023年11月9日(星期四) 下午5:05 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [FreyrS/dMaSIF] why the interface is computed like this ? (Issue #42)

Hi @camel2000, @YAndrewL ,

Apologies for the long delay. To answer briefly: please note that the .step() function is "old" KeOps syntax that means "> 0". As a consequence, writing: xyz_dists = ((xyz1_i - xyz2_j) ** 2).sum(-1).sqrt() xyz_dists = (1.0 - xyz_dists).step()

creates a (N_1, N_2) LazyTensor of "booleans" whose value xyz_dists[i, j] indicates if point i from protein 1 and point j from protein 2 are at distance < 1 Angstrom from each other. Then, computing: p1_iface_labels = (xyz_dists.sum(1) > 1.0).float().view(-1)

is just a way of looking for points i in protein 1 with at least two such neighbours in protein 2. All in all, this code "flags" points on the surface of protein 1 that are close to the interface of protein 2, and vice versa.

I hope that this answers your question, please feel free to re-open the issue if needed :-) Best regards, Jean

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>