search

zaixizhang / MGSSL

Official implementation of NeurIPS'21 paper"Motif-based Graph Self-Supervised Learning for Molecular Property Prediction"
Apache License 2.0
111 stars 16 forks source link

About Motif-Decomposition #9

Closed JongKook-Heo closed 1 year ago

JongKook-Heo commented 1 year ago

Hi zaixizhang, Thanks for sharing your nice works!

I have some questions about generating cliques and bonds in your code.

  1. Is there any reason to set MST_MAX_WEIGHT to 100?(motif_based_pretrain > util > chemutils.py) 1-1. If not, how does the MST_MAX_WEIGHT effect to generated cliques?(i.e. If MST_MAX_WEIGHT raises, it causes almost infinite number of motifs?)

  2. In MolTree, When the number of edges connecting BRICS cliques is less then 1, Tree decomp overwrites cliques and edges. Does it okay to change the threshold 1 to any small integer like 2 or 3?

Sincerely, JongKook Heo

zaixizhang commented 1 year ago

Hi JongKook,

Thanks for your interest in our work and the great questions!

  1. The set of MST_MAX_WEIGHT has little influence on the generated cliques. As shown in chemutils.py, we assign smaller weights to edges between cliques/motifs with more atom intersections (MST_MAX_WEIGHT - number of shared atoms). Then we use the minimum_spanning_tree function to select the edges between cliques.
  2. Yes, you can change the threshold to other small integers. Generally, tree decomp provides finer molecule decompositions.