RosettaLigand: How to score native pose in protein-ligand docking experiment?

[Anonymous]

Hi Everyone:

I’m working on protein-ligand docking experiment. I have an odd problem concerning how to calculate the “native pose” total_score , interface_delta and ligand_auto_rms_no_super value from rosettaligand docking experiment.

I have getted silent file after finishing rosettaligand docking experiment, and I had analysised output silent file with that contains the endpoint of every trajectory. I had also plotted interface_delta for these top-scoring structures against ligand_auto_rms_no_super and I had get a nice docking funnel. But I don’t know how to calculate the “native pose” total_score , interface_delta and ligand_auto_rms_no_super value for comparing the predicting complexes. The only predicting structures docking funnel is thin for my paper. In 2008, Ian W. Davis and David Baker. “ROSETTALIGAND docking with full ligand and receptor flexibility” Fig.2.(C) had calculated “min. native”, but I don’t kown how to get it.

Maybe you know how to calculate it. so I hear you suggestions.

Thank you in advance.

Ryhon Wang

11.10.2015

[Anonymous]

If there is a native structure availible, there are options where you can turn off the global sampling during ligand docking. If you’re using the newer XML based protocols, you’d want to remove any SlideTogether/Transform/Translate/Rotate/HighResDocker steps, and only use the FinalMinimizer step. If you’re using the older  application-based ligand docking, then there’s a “-docking:ligand:protocol” option which you can set to “min_only”, which will do just the final minimization step of docking, and not the global sampling.

Of course, much of the time when you’re actually using docking (as opposed to making a benchmarking paper) you won’t have the native information. In those cases, a “min. native” entry really wouldn’t apply, and you can just skip the calculation.

[Anonymous]

One more question!

I run rosettaligand using original flags, only the “-docking:ligand:protocol” option configured to “min_only”, and nstruct option is configured to 20. After that, I got the silent files, total_score , interface_delta and ligand_auto_rms_no_super values about 20 near native complexes. These near native complexes have a better ligand_auto_rms_no_super values, but the interface_delta values are bad. So the figure is not nice for my paper. I don’t know why these near native complexes have bad interface_delta values simultaneously get a nice ligand_auto_rms_no_super values.

the flags file is below:

-in

 -path

  -database /home/sky/rosetta3.4/rosetta_database

 -file

  -extra_res_fa ./inputs/1AQ1_A_L_Conformers/1AQ1_A_L.params

  -s ./inputs/1AQ1_A_Pose_MiniEnergy/1AQ1_A_Pose_input.pdb

  -native ./1AQ1_A_NoWater.pdb

 -detect_disulf false

-out

 -mute protocols.geometry.RB_geometry core.scoring.dunbrack.SingleLigandRotamerLibrary core.scoring.rms_util

  -nstruct 20

  -file

  -renumber_pdb

 -path

  -pdb ./outputs/

-run

 -rebuild_disulf false

 -version

 -packing

 -ex1

 -ex1aro

 -ex2

 -extrachi_cutoff 1

-docking

 -uniform_trans 5

 #-randomize2

 #-random_conformer

 #-dock_pert 30 5

 -ligand

  #-improve_orientation 1000

  -minimize_ligand

  #-harmonic_torsions 10

  #-minimize_backbone

  #-harmonic_Calphas 0.3

  #-soft_rep

  #-old_estat

  #-protocol abbrev2

  #-start_from  -1.731  32.589  -5.039

  #-mutate_same_name3

#-mute core.util.prof ## dont show timing info

#-mute core.io.database

And score.tab critical columns about 

SCORES angle_constraint interface_delta ligand_auto_rms_no_super ligand_auto_rms_with_super total_score

1AQ1_A_Pose_input_0001 0 44.5899 6.56051 1.26391 -706.56

1AQ1_A_Pose_input_0002 0 523 6.56868 1.17867 -213.506

1AQ1_A_Pose_input_0003 0 21.4006 5.42768 1.23106 -756.455

1AQ1_A_Pose_input_0004 0 7.14631 5.37769 1.20015 -787.577

1AQ1_A_Pose_input_0005 0 -2.38131 7.88429 1.21405 -803.479

1AQ1_A_Pose_input_0006 0 0.719705 6.17194 1.20776 -817.568

1AQ1_A_Pose_input_0007 0 6.92893 7.37052 1.20098 -803.847

1AQ1_A_Pose_input_0008 0 13.8084 9.03073 1.2121 -788.857

1AQ1_A_Pose_input_0009 0 1.21918 10.3068 1.2059 -810.712

1AQ1_A_Pose_input_0010 0 -5.25537 10.4907 1.20024 -820.575

1AQ1_A_Pose_input_0011 0 17.4131 5.88176 1.22542 -773.475

1AQ1_A_Pose_input_0012 0 56.6036 6.56316 1.24887 -632.044

1AQ1_A_Pose_input_0013 0 1.74905 6.14208 1.21873 -804.52

1AQ1_A_Pose_input_0014 0 229.752 6.16497 1.20006 -543.859

1AQ1_A_Pose_input_0015 0 245.178 6.15812 1.21037 -496.522

1AQ1_A_Pose_input_0016 0 24.3403 5.28721 1.19788 -767.805

1AQ1_A_Pose_input_0017 0 327.492 5.36994 1.24199 -397.743

1AQ1_A_Pose_input_0018 0 11.2781 5.46825 1.22278 -759.583

1AQ1_A_Pose_input_0019 0 1.65718 6.46323 1.20701 -803.446

1AQ1_A_Pose_input_0020 0 5.48646 5.36094 1.22061 -777.239

 

In 20 near native complexes, the best interface_delta value is  only -5.25537, it is a not very good value. But the ligand_auto_rms_no_super values are better, and the values are near 1.2A.

This is an realy odd problem!!!

Maybe you know why happend to it.  I hear you suggestions.

Thank you in advance.

Ryhon Wang

11.11.2015

[Anonymous]

Thanks for rmoretti’s advices.

By the way, I am using rosetta3.4 version. So I will try “min_only” option.

[Anonymous]

min_only giving you better ligand_auto_rms_no_super  is not surprising. Since you’re not doing active sampling of the ligand position, your not going to move as far from the starting position as you would with a full sampling protocol.

The worse interface_delta score is a bit more concerning, as one would hope that the native complex would have the best energies. But it could be that Rosetta thinks the alternate binding conformation is the better binding one. That is not unheard of. One thing you might want to think about is where your native reference is coming from and if you trust the location of the ligand.

Another issue that might be coming up is due to pre-relaxing the protein, or lack thereof. min_only does minimization, and if slightly more agressive optimization is needed to optimize the ligand (e.g. the sidechain rotamers in the active site are not in the best of locations), then things won’t work to well. 

Digging a bit, I think I’ve come across the recommended options for the Davis & Baker 2009 J. Mol. Biol paper for changing a docking run into a “native” run, and it looks like it’s as simple as removing any gross sampling options like -uniform_trans;-improve_orientation; -randomize2; etc. You still keep the “-protocol abbrev2” to do local repacking and minimization around the ligand. It’s also recommended that you add in the “-use_input_sc” flag, to allow the protocol to take advantage of the native ligand conformation.

    

[Anonymous]

Thanks a lot for your advice in detial, rmoretti!

[Author]

Posts