I would like to know the meaning represented by each dimension of the contact map, distance map and HB map.

[Liuyu_20000419]

I read the manuscript "Deep-learning-based single-domain and multidomain protein structure prediction with D-I-TASSER ", where Fig 1a shows the contact map, the distance map and the hydrogen bond network map, I am confused about the dimensions of these three. According to my understanding, these three graphs can be represented by the matrix of L * L. May I ask what the meaning of "2", "3" and "N" here respectively? Why the last dimension of the distance map and hydrogen bond network map are "N" but the last dimension of the contact graph is "1"? And why does the hydrogen bond network have an extra L dimension?
Looking forward to your reply, thank you!

[jlspzw]

Dear user,

Thank you for raising this question. We have two kinds of contacts, contacts between CA atoms, and contacts between CB atoms, so in the contact map figure, we have 2 before LxL, similar to the distance map, we have CA distance and CB distance. For distance prediction, we did not directly predict the distance between two residues; instead, we predicted the distance drop into which distance intervals, like 1-2A, 2-3A, etc, so the N means the number of the intervals. For hydrogen bond networks, we have AA, BB, CC hydrogens, and similar to distances, we predict the angles drop into intervals, so we should have the size 3xLxLxN., I think we have a typo in the figure; we do not have the third L there.

Thank you.

Best Regards
Wei Zheng

[Liuyu_20000419]

Dear user,

Thank you for raising this question. We have two kinds of contacts, contacts between CA atoms, and contacts between CB atoms, so in the contact map figure, we have 2 before LxL, similar to the distance map, we have CA distance and CB distance. For distance prediction, we did not directly predict the distance between two residues; instead, we predicted the distance drop into which distance intervals, like 1-2A, 2-3A, etc, so the N means the number of the intervals. For hydrogen bond networks, we have AA, BB, CC hydrogens, and similar to distances, we predict the angles drop into intervals, so we should have the size 3xLxLxN., I think we have a typo in the figure; we do not have the third L there.

Thank you.

Best Regards
Wei Zheng

I get it. Thank you for your patient explanation！

[Xerianna]

I read the manuscript "Deep-learning-based single-domain and multidomain protein structure prediction with D-I-TASSER ", where Fig 1a shows the contact map, the distance map and the hydrogen bond network map, I am confused about the dimensions of these three. According to my understanding, these three graphs can be represented by the matrix of L * L. May I ask what the meaning of "2", "3" and "N" here respectively? Why the last dimension of the distance map and hydrogen bond network map are "N" but the last dimension of the contact graph is "1"? And why does the hydrogen bond network have an extra L dimension?
Looking forward to your reply, thank you!

The contact map uses a final dimension of 1 because it only encodes binary information Poor Bunny —whether a contact exists between residues.

The distance map and hydrogen bond network map have a final dimension of N, which typically indicates multiple distance bins or types of interactions being encoded simultaneously.

The hydrogen bond network may include an extra L dimension (resulting in L × L × L × N or similar) to capture higher-order or directional dependencies, such as donor-acceptor relationships involving a third residue.