Ligands in Uploaded Protein Files Are Kept in Structure #60
Description
Hi,
I'm opening this issue to document a problem we observed when uploading protein files that include a ligand in the pocket. When you upload a protein file with the protein bound to a reference ligand, the ligand is kept in the structure. This is not immediately noticeable in the WebApp because the default structure visualization is set to 'cartoon'. However, when we switch to the 'sticks' view, we can clearly see that the ligand is still in the pocket. Below is a screenshot showing the input structure (light blue) and an example generated ligand (gray), where we see that the ligand from the input protein file is still present.
The observed effect is the generation of (PoseBusters) invalid structures with large ring systems, caused by an overcrowded pocket. Likely, the ligand is being treated as an NPNDE, but this should be confirmed. Some examples of generated ligands are shown below.
This is not immediately fixable by removing all non-protein atoms/structures because we may want to model non-protein structures such as cofactors. Potential fixes for the WebApp may include an option to select or deselect non-protein structures for inclusion. We could also simply display a warning if the pocket seems to be occupied by non-protein atoms, which is likely the only option for the CLI.