search

jasonkyuyim / se3_diffusion

Implementation for SE(3) diffusion model with application to protein backbone generation
https://arxiv.org/abs/2302.02277
MIT License
320 stars 51 forks source link

How to generate dynamic images in the demo? #28

Closed wewewexiao2008 closed 1 year ago

wewewexiao2008 commented 1 year ago

The initial secondary structure exhibits remarkable characteristics. I'm wondering how to reproduce it with my own structure.

jasonkyuyim commented 1 year ago

Hi, I use https://marketplace.visualstudio.com/items?itemName=ArianJamasb.protein-viewer

Take x0_traj_1.pdb from a good looking sample after running inference (I use noise scale 0.1) then open up protein-viewer on this file.

  1. Click Stylized in "Quick styles".
  2. Change "Set coloring" to Atom property -> Element Index.
  3. Export animation where there should be a button to do the reverse trajectory.

Attached is a screenshot of what the UI looks like for me.

Screenshot 2023-07-22 at 12 03 52 AM
jasonkyuyim commented 1 year ago

https://github.com/jasonkyuyim/se3_diffusion/assets/74610237/51e1099f-2004-44f8-9c2f-95c2702eea7d

wewewexiao2008 commented 1 year ago

Thanks a lot! Another question: It has come to my attention that despite the presence of multiple models in the pdb file, the model number consistently appears as 1 instead of 50, or 500, as observed in your case. The RFDiffusion is employed for my model generation. Could you kindly provide information regarding the specific label that requires modification in the pdb file to facilitate successful parsing by pdbviewer? image

wewewexiao2008 commented 1 year ago

The annotation for MODEL 1 and MODEL 2 is missing in the PDB file. After adding them, the recognition can be performed correctly.

image