Ambiguous restraints random choice

HADDOCK
1

The documentation mentions that by default 50% of ambiguous restraints is chosen from the input file. I have a few questions:

  1. Does this random 50% changes for every pose or it stays the same, so poses differ only by random translation/rotation at initialization.
  2. Does this choice change between it0 and it1 for a particular pose? (assuming the same ambiguous restraints file is present in both stages)
  3. How to find out which 50% of restraints have been chosen?
  4. Is 0.01 weight for Eair in scoring function enough to drive the docking towards distances in restraints? How this weight was chosen? As various terms (EVdv, Eelec, Edesolv, Eair) may not be directly comparable by value, how much is the pose actually penalized when not fulfilling the distance in restraint? Is there any way to penalize it even more?

Thank you!

2
  1. Does this random 50% changes for every pose or it stays the same, so poses differ only by random translation/rotation at initialization.

For every pose a different 50% is used.

  1. Does this choice change between it0 and it1 for a particular pose? (assuming the same ambiguous restraints file is present in both stages)

No the same set is used for it0 and it1 (which is why you will find .seed files created in it0 and it1)

  1. How to find out which 50% of restraints have been chosen?

You can’t :slight_smile:

There is however an alternative: You could create your own set of restraints, one for each model to be generated at it0. Name the files as ambig.tbl_X where X is the model number and manually place those files in the it0 directory. If present this file will be used. noecv should in that case be set to false

  1. Is 0.01 weight for Eair in scoring function enough to drive the docking towards distances in restraints? How this weight was chosen? As various terms (EVdv, Eelec, Edesolv, Eair) may not be directly comparable by value, how much is the pose actually penalized when not fulfilling the distance in restraint? Is there any way to penalize it even more?

The weights used in scoring are not those used during the calculations.

See: https://www.bonvinlab.org/software/haddock2.4/run/#distance-restraints

In it0, the distance restraint weight is slowly increase. If you want to know how search for Ambig_scale in refine.inp

The Eair at it0 is usually quite larger than the Evdw and Eelec contributions, which is why its weight is set to 0.01.But you can of course play with those.

The current weights have been doing a good job on average over more than 17 years by now :slight_smile:

1 Like