I am trying to build a model where the clash score is around 2.0 before ligand fitting. However increasing drastically after ligand fitting (>30). It is a cryo-EM map of ~3.2 A resolution with three ligands (very clear density for all). Out of three one ligand is showing density for the alternate conformer and one ligand is new where I am generating the restrain file using ArcDRG tool in ccp4i. I tried to fit the model in deepEMhancer map and the refined map both. Both are giving the same issue. I am using coot for the model building and phenix for refinement. Could you please help?
This is the CryoSparc forum, you’ll probably have better answers if you post this in CCPEM or 3DEM.
That said, my tips for improving clash scores:
add Hydrogens before refinement (Reduce, Phenix…) - try different ones, this is curious but different programs will give different positions for hydrogens.
Always visually check hydrogen bonding of all Gln, Asn, Thr, His after adding hydrogens, switch rotamers if necessary - don’t blindly trust the programs on this.
try molecular dynamics-based tools e.g. ISOLDE for refinement.
Also check if your ligands are really intact in the sites. Some of them might degrade in the tube.
Is the clashing caused by ligand with protein or self ligand-ligand clashing (view the molprobity chart in phenix validation or click phenix’s open in coot and view the clashing)? With proper restraints, refining in coot followed by refmac might help. Phenix refinement tends to make some things worse.
Clashing might indicate heterogeneity. How do the ligands and surrounding region look in local filtered map? Is the resolution lower than other high res regions?
3D classification on ligands with surrounding residues might enhance resolution/ligand density if it’s a heterogeneity problem.
I am wary of deepEMhancer maps as they can remove/modify the ligand density.