Repository navigation
Question regarding poses.csv output #6
Description
I ran combind and received the following output for poses.csv
ID,POSE,COMBIND_RMSD,GLIDE_RMSD,BEST_RMSD
ligand_6,0,0.883745269850613,0.883745269850613,0.22876830274335147
ligand_7,0,-1.0,-1.0,-1.0
ligand_8,1,-1.0,-1.0,-1.0
ligand_9,0,-1.0,-1.0,-1.0
ligand_a,0,-1.0,-1.0,-1.0
ligand_b,0,-1.0,-1.0,-1.0
Should I be concerned that only one ligand has RMSD's that are not -1? Can you help me understand what this output means. Thank you!
Activity
jpaggi commentedon May 12, 2022
Hi Hunter, the POSE column gives the index of the pose that was selected by combind and the remaining RMSD columns provide the RMSD to the native pose of the combind pose, glide pose, and the best (lowest-rmsd) pose in the pose list.
An RMSD of -1 reflects that no native structure was provided for that ligand in the "combind featurize" step. So having many -1's is common. If you thought that you had provided native poses for the other ligands, there is probably a mismatch between the file names of the native poses and the entry titles in the docking results.
Hope that helps!
hnisonoff commentedon May 13, 2022
Thank you! And should I interpret pose 0 as meaning that Glide would have selected these as the top pose without the combind correction factor?
jpaggi commentedon May 15, 2022
Yes exactly. In this example, the poses for all but ligand_8 are unchanged from the Glide prediction.
hnisonoff commentedon May 16, 2022
Great, thank you for all the quick replies! I really enjoyed the paper!
hnisonoff commentedon Jun 11, 2022
@jpaggi one other question. In the SI of the paper it sounded like you only use the shape similarity features for virtual screening, but I noticed in the code and README that shape similarity is used in addition to MCSS similarity for normal docking pose prediction. Do you have any guidance for whether I should be using both or just one? Can I ignore shape similarity if I am using MCSS?