Visualizing crystal contacts with PDBe-Molstar

A great gotcha in structure-based hit discovery is when a binding site or bound ligand is affected by crystal contacts. Crystal contacts here refers to adjacent protein molecules contacting each other in the crystal used for X-ray crystallography in a way that would not happen in solution. While having some contacts is normal, they become problematic when they distort the ligand or surrounding protein.

For example, take a look at the binding mode of BI-2852, a KRAS inhibitor that came out of a fragment screen. Interestingly, this arose from a 2 milliMolar, fragment. It was optimized to a sub-microMolar binder, despite not making many more protein contacts. Half of it is barely hanging on to the protein yet, for some reason, it is stable in this position as judged by it's well-resolved density (per the validation report on RCSB).

This odd ligand behaviour is a candidate for crystal contact interference. We can use PDBe-Molstar, a convenient protein+ligand visualization tool, to see these. The key line is setting the option defaultPreset to 'unitcell' - see an example below and inspect the source of this page to see how that works. You'll notice that the BI-2852 molecules sit snugly between two copies of KRAS, explaining the stability of the apparently solvent-exposed groups. Click one of the intercalating ligands and you'll see that about 50% of the buried surface area is in an adjacent protein unit.

Visualization:

It's arguable whether the crystal contacts are relevant in this case. Indeed some researchers argue they are - and the BI folks argue they aren't! Having a tool to visualize these at least helps one to make a judgement themselves.

Alternatives: I often just open a pdbe-molstar plunker, change the pdb code, and add defaultCell: 'unitPreset'. That isn't secure, and requires internet. PyMol has a symexp command if you can bare to use it, and ChimeraX has a crystal contacts tool as well, but the resulting visualization can churn through CPU.