amino acid frame triplet side chain
Closed this issue · 3 comments
Hello, thanks for open sourcing your project.
I was wondering if I train ScanNet from scratch, should I modify line 170 to use Nitrogen atom of the current amino acid atom_coordinate
instead of the previous aimo acid i.e. atom_coordinates[l-1]
. I noticed in your comment that it was due to a previou issue. So just to confirm if I retrain the network, my understanding is I should replace line 169-175 line 182-188, is that correct?
ScanNet/preprocessing/protein_frames.py
Lines 166 to 188 in 77d8510
Hi,
Thanks for your interest in our research.
Regarding protein_frames.py, please leave as it is, the change was already made.
However, if you retrain, you can comment l427 of file preprocessing/protein_chemistry.py and uncomment l428.
This is a bug regarding histidine parsing that I only noticed recently.
Best regards,
Jérôme
Thansk for your reply! Though I am still confused about line 169-171 in protein_frames.py
For examle, a Glycine (residue index l
), as in line 169-171 you are calculating its virtual side-chain CoM using the glycine's C and Cα atom coordinates and the Nitrogen atom of its preceding residue (l-1
) if I understand correctly. In your paper, the side chain CoM was defined using Cα
, C
, N
atom vector coordinates of the residue. So, why using nitrogen from preceding residue atom_coordinates[l-1][atom_ids[l-1].index(17)]
rather than the glycine its own nitrogen atom e.g. atom_coordinate[atom.index(17)]
?