Apo targets not getting pocket center correctly defined

[j-wags]

From email: We also want to report a potential problem regarding apo structures that CELPP provided. For a few targets in recent weeks, apo structures were not mapped to other protein structures (e.g. LMCSS). In these cases, predictions would be failed if the binding site information in the file "center.txt" was used for docking. The following are a list of problem targets that we found. Hope these are helpful.

week22: 5lzj, 5lzg (celpp 1.8.0)

week 23: 5vt4, 5x8s (celpp 1.8.0)

week 24: 5suk (celpp 1.8.0)

week26: 5val, 5vam, 5vyj, 5xai (celpp 1.9.0)

[j-wags]

My first suspicion is that we are somehow forgetting to do an alignment for the apo candidate. This timeline suggests that perhaps the change was when we started doing binding site alignments, which would fail for apo as it doesn't have a ligand to define a binding site.

[j-wags]

Fixed by commit bf15a43 to master (not tagged/deployed)

[j-wags]

Pushed to staging. Need to ensure that this works correctly when CELPP runs week 32.

[j-wags]

I've checked the week 32 challenge package for cases where this failure occurred on the celpp1 server. These failures are:
5gqe
5myg
5vyg (this is a really strange case)

On celpp-stage, 5gqe and 5myg are fixed. Therefore I think this bug is fixed and the code can be pushed to production.

5vyg is a different case and I'll open up an issue ticket for it.