devalab / DeepPocket

Ligand Binding Site detection using Deep Learning
MIT License
89 stars 26 forks source link

Channels in training script are different from those in your Supporting Information table s1 #24

Closed ckami12138 closed 1 year ago

ckami12138 commented 1 year ago

Thanks for your great work. I have one quesion, in your training code, you use the first 14 channels in gninamap file, which are

Hydrogen, PolarHydrogen, AliphaticCarbonXSHydrophobe, AliphaticCarbonXSNonHydrophobe, AromaticCarbonXSHydrophobe, AromaticCarbonXSNonHydrophobe, Nitrogen, NitrogenXSDonor, NitrogenXSDonorAcceptor, NitrogenXSAcceptor, Oxygen, OxygenXSDonor, OxygenXSDonorAcceptor and OxygenXSAcceptor

, However in Supporting Information table s1(https://pubs.acs.org/doi/suppl/10.1021/acs.jcim.1c00799/suppl_file/ci1c00799_si_001.pdf), they are

AliphaticCarbonXSHydrophobe, AliphaticCarbonXSNonHydrophobe, AromaticCarbonXSHydrophobe, AromaticCarbonXSNonHydrophobe, Bromine Iodine Chlorine Fluorine, Nitrogen NitrogenXSAcceptor, NitrogenXSDonor NitrogenXSDonorAcceptor, Oxygen OxygenXSAcceptor, OxygenXSDonorAcceptor OxygenXSDonor, Sulfur SulfurAcceptor, Phosphorus, Calcium, Zinc and GenericMetal Boron Manganese Magnesium Iron

why they are different, may I misunderstand something?

RishalAggarwal commented 1 year ago

Yes, gninamap file is used to create the gninatype files which specify the atom type a particular atom is. When these files are loaded into libmolgrid, libmolgrid assigns these atoms to different channels based on whether it is a "receptor" or a "ligand" atom and based on the atom type identity. The first 14 channels represent the receptor atoms - theres probably a better way to do this but its a hack that worked at the time. So the first 14 channels represent the atom types that are given in our supporting information.

ckami12138 commented 1 year ago

Thank you for your quick response. So, you mean libmorgrid considers Bromine Iodine Chlorine Fluorine (which are 4 diferent channels in gninamap file) as channel 5 (number from 1), Nitrogen NitrogenXSAcceptor as channel 6 (which are 2 diferent channels in gninamap file) etc.. Is this right? So, what do the other 14 channels represent, and where can I find the definition of the whole 28 channes in libmolgrid files?

RishalAggarwal commented 1 year ago

you can see the ligand atom types here: https://github.com/gnina/libmolgrid/blob/master/include/libmolgrid/atom_typer.h - the atom types are at the bottom of the file

ckami12138 commented 1 year ago

So, does it mean that if I use the defaultGninaReceptorTyper, I wiil get the same representation of protein as in your paper?

RishalAggarwal commented 1 year ago

Yes you should when you load a grid with it, however you have to be careful not to save gninatypes files using that typer

ckami12138 commented 1 year ago

Thank you very much for your patience. My doubts have been resolved. I will close this issue.