Energy Minimization To Determine Backbone Clashes

Question

[Harpalss]

I am currently modelling a protein, I have so far generated 160 models of this protein.

I then remve models that have backbone clashes via energy minimisation. Models with a -ve energy after minimisation were considered to have no clashes where as proteins that have +ve energy after minimisation were considered to have a clash.

I then looked at structures that had -ve energy and some of them had severe backbone clashes, where as other model with a +ve energy did not.

Is there any reason for this? Which is the best way (rather than eye) to determine if a protein has a clash?

I performed minimisation in Chimera. I did 100 steps, with 0.02 angstrom movement and no fixed atoms.

Also, is it better to look at the energy or the gradient norm? I used energy because I do not know what the gradient norm is.

Answer 1

[João Rodrigues]

How are you modelling the protein? I am surprised that a lot of people use Chimera and Pymol to model proteins when their function is strictly graphical.. If you are modelling something, I would recommend you to use MODELLER, or I-Tasser if homology is low.

Well, energy minimization MIGHT get rid of all the clashes, but it depends on how strong the minimizer is and for how long it runs. Download GROMACS, install it, and run energy minimization until machine convergence. That is much longer than 100 steps and might fix more clashes. Also, use WHATCHECK or another software like that to check for stereochemistry of your models instead of just looking for clashes.