How do the concept wiki free text branch filter options map to 'things' in the linked data?
The search results will be filtered by this source authority; optional. Possible values: 1 = Community, 2 = UMLS, 3 = SwissProt, 4 = ChemSpider, 5 = Computer Inferred, 6 = Pathway Ontology, 7 = WikiPathways
For example, which filter (if any) would return chembl target components only? Is it possible to ask only for compounds which come from OCRS or drugbank etc?
What do 'Community', 'UMLS' and 'Computer Inferred' mean?
These select the data sources that were used to get the labels to ConceptWiki. ConceptWiki then often merged entities of two different datasources into the same concept. E.g. if you have a look at http://www.conceptwiki.org/concept/index/9bc59ebd-9e3a-47b2-a179-547efb6a66b8, you can see the labels from Chemspider and from UMLS (Unified Medical Language System). Chemspider usually has more synonyms, but from UMLS you would get a description of the molecule. In Open PHACTS, only the ConceptWiki molecules which are from ChemSpider are then mapped to other data sources, so if you want to search for a molecule, branch 4 would be the option to choose.
Similarly, for proteins you have SwissProt and UMLS as datasource, e.g. http://www.conceptwiki.org/concept/index/6e727c77-69ac-40f2-9754-34b930bb3200. Here, SwissProt would be the main option to choose for proteins, as these concepts are more likely to retrieve data. But you could use a UMLS Concept as a tag to find proteins of this class, e.g. http://www.conceptwiki.org/concept/index/c0cad170-db0a-488e-a3ab-a892a4fa4a55.
There is, however, no filter for ChEMBL target, as the ChEMBL database was not used as a source for synonyms. Also OCRS and drugbank are not included here.
The Community names at the top were initially planned to allow edits from the users, but this is currently not implemented. I think these are currently a combination of the other datasources.
UMLS is a classification for medical terms. You'll find a term there for many compounds (especially drugs), proteins, diseases and so on. See also http://www.nlm.nih.gov/research/umls/.
Computer Inferred are concepts which were created, for example by removing all the species information from the name. I think this was done by removing everything that was in brackets. You can then use this concept as a tag, to find all concepts which have this name in common.