Setting up an alert for new human PDB ligands

Question

[cdsouthan]

Our database team (http://www.guidetopharmacology.org/) would like to set up some kind of allerting triage for the following criteria for human proteins a) novel ligands in an old protein b) old ligands in different proteins b) a protein that gets any ligand for the first time. There are some additonal filters such chemical property and pocket locations indiciative of specifically bound, lead-like, small-ish, moleculular structure (i.e.pharmcologically-relevant ligands rather than just new hetero atom enties). We'd also want a frequent-hitters stop list. For example, we're not that interested in seeing yet more ligands from popular kinases, BACE1, DPPIV, FX or thrombin (exept for clinical cpds of course but we can select these anyway).

None of the three WWPDB portals seem to allow such a query off the bat. In addition we would prefer a transatlantic split in the form of Swiss-Prot IDs for the proteins but PubChem CIDs for the ligands (i.e. rather than gi numbers or UniChem IDs but we could cope with these). For the record we do check MMDB for novel CIDs and skim the PDBe weekly new entry lists but a specific alert is obviously more efficient. Might a SPARQL query or a KNIME node be an option?

[cdsouthan]

This is still a gap for us that we can't really fill with a KNIME node or local PDB copy just now. Anyone know any other ways forward (BioMart?)

Answer 1

[pwrose.ucsd]

The folks at Vernalis have written KNIME nodes that access [RCSB PDB web services].

Check out this presentation how you might use KNIME to create an alert service for new ligands.

[cdsouthan]

Useful thanks

Answer 2

[João Rodrigues]

I think it's better to have a local copy of the PDB updated in your servers and everytime there is an update, check what is coming in. You should build a database with the information you are looking for (who has ligand, which proteins have which ligands, etc) and then just compare. It's a one-off effort to build this database but then it's only about incremental updates and checks, which could make it very manageable. You can even get rid of the structural data once you populate the DB to save space.

p.s. I'm not very familiar with KNIME nodes or SPARQL queries, so apologies for the general answer..

[cdsouthan]

Thanks, we'll cogitate on this

Answer 3

[conroy]

This URL http://www.ebi.ac.uk/pdbe/entry/search/index?document_type:latest_chemistry&view=compounds&organism_scientific_name:%22Homo%20sapiens%22 will display the ligands which are new to the PDB in the current week which are bound to human proteins

[cdsouthan]

Ta, it works - super