Are the omega, theta and phi angles corresponding to the main chain torsion angles omega, psi and phi angles?

Question

Are the omega, theta and phi angles corresponding to the main chain torsion angles omega, psi and phi angles?

Wenhao-Jin opened this issue 3 years ago · 7 comments

Hi authors, I am trying to use your MIF-ST model and now preparing the node/edge features. Just to be sure I got the features correctly, I was wondering if the omega, theta and phi angles in your paper are referring to the torsion angles that are used in peptide/protein structure (definitions can be found here).

According to Figure 2 in your MIF paper, the answer is no. But the Figure 2 was a bit confusing to me since the omega, theta and phi are all related to the beta carbons. However, not all of the amino acids have beta carbons. I also checked the paper from Ingraham et al. (2019) which is cited in your MIF paper for this part. In that paper they didn't give much details but used the terms "omega", "psi" and "phi". Could you help me with this? Great thanks in advance!

Answer 1 · 2022-06-25T13:22:04.000Z

Yes, they are the standard definitions. We use the same method for calculating them from structure as Ingraham. Thanks for pointing out that the figure is wrong: I'll work on fixing that.

…

On Sat, Jun 25, 2022, 4:50 AM Wenhao-Jin ***@***.***> wrote: Hi authors, I am trying to use your MIF-ST model and now preparing the node/edge features. Just to be sure I got the features correctly, I was wondering if the omega, theta and phi angles in your paper are referring to the torsion angles that are used in peptide/protein structure (definitions can be found here <https://swissmodel.expasy.org/course/text/chapter1.htm> ). According to Figure 2 in your MIF paper <https://www.biorxiv.org/content/10.1101/2022.05.25.493516v1.full.pdf>, the answer is no. But the Figure 2 was a bit confusing to me since the omega, theta and phi are all related to the beta carbons. However, not all of the amino acids have beta carbons. I also checked the paper from Ingraham et al. (2019) which is cited in your MIF paper for this part. In that paper they didn't give much details but used the terms "omega", "psi" and "phi". Could you help me with this? Great thanks in advance! — Reply to this email directly, view it on GitHub <#7>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADEMNWB6M7C6MWMOFDCKY3LVQ3B67ANCNFSM5ZZ6LKHA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

Answer 2 · 2022-06-25T19:05:52.000Z

Got you. Thanks for clarifying it, @yangkky!

Answer 3 · 2022-06-26T02:36:25.000Z

@Wenhao-Jin Just wanted to clarify on this - the angle features used follows the angles described by the trRosetta paper and unconventional compared to how dihedral angles for proteins are typically defined.

Answer 4 · 2022-06-26T07:04:48.000Z

@hyeh20 Thank you so much for the clarification! Just to make sure, for the dist value, I was also wondering if MIF also used the distance between beta carbons as trRosetta did. Besides, is there some script or package available for generating these angles. Thanks so much again!

Answer 5 · 2022-06-26T19:21:58.000Z

Yes, the distance should be between beta carbons, we will update our figure in our next revision.

For scripts on how to generate these angles: If you are starting with a pdb file for your proteins, please use the functions in pdb_utils.py. parse_PDB will generate a coordinates dictionary and then you can use process_coords to generate the angle features. These scripts were directly modified from the trRosetta paper.

Answer 6 · 2022-06-27T14:50:23.000Z

@Wenhao-Jin sorry for the confusion: @hyeh20 is correct.

The parsing functions are here: https://github.com/microsoft/protein-sequence-models/blob/main/sequence_models/pdb_utils.py

Answer 7 · 2022-09-16T19:54:08.000Z

Thanks so much for the help! Sorry for the delayed response, it works on my side. Really appreciate it!