search

wyji001 / Point-Cloud

20 stars 6 forks source link

Question about preprocess the PDBbind-2016 database #4

Open SharplessSword opened 2 years ago

SharplessSword commented 2 years ago

Hi, after a long time, i think i can train my PointNet on the refined-dataset. But as you said, the indicator is after a manner.So i turn my eyes to bigger dataset. In the paper , you describe as (1) Entries that contain peptide ligands were deleted (590); (2) The entries containing covalently-bonded ligands were deleted (379); (3) The entries with insufficient ligands were deleted (481). I'm not sure how to do it the right way(especially the second rule), could you give a more detailed guide? Thanks again!

SharplessSword commented 2 years ago

@wyji001

wyji001 commented 2 years ago

PDBbind already provides enough information, you just need to filter them out.

SharplessSword commented 2 years ago

@wyji001 sorry, i'm new to this field, and not familiar PDBbind. Could you tell how to filter them out, I skim the pdbbind-2016 datatset, which i download, the most part of the data is the coordinates. Such as, how can i judge an Entreis weather contains peptide ligands or not, how can i judge an Entreis weather containing covalently-bonded ligands or not, how can i judge the entries weather contains insufficient ligands.

wyji001 commented 2 years ago

This file "INDEX_general_PL.2016".