Semantic resources project/Meeting notes/2010-02-11

Attendance:

TD, PC, AR, JAR.

Outline:

- Discussion of PRO OBO file from Darren -- TD: shows EW's questions. -- Discussion of "only_in_taxon" -- Protein complexes are represented, at least in nonce form, in the file we received. - TD: Talks about name->protein mapping -- TD: demo of interactive (Lucene) search -- TD: HSP submission table. --- A lot of the columns here are redundant: species? (but this is already in UNIPROT). --- Protein object name is the same issue. --- Is the evidence code necessary when we've already got the UNIPROT? --- In uniprot, they use the word "fragment" to say that it's not complete. -- TD: SWAN Proteins --- 100+ proteins were identified from SWAN --- All but ~10 were already listed in PRO --- (But not in species-specific form) --- three human/mouse proteins weren't in the list at all -- of these, one was already in Paolo's list (HSP104)

- Antibody modeling -- PC: TD compiled a list of terms from OBI to use in modeling. -- PC shows his slides of "A Manual Annotation Story of Beta-APP LN27". -- TD: shows the preliminary OWL file in protege -- Discussion of evidence codes for representing provenance and confidence: --- AR: tended to use owl annotations as a way of asserting evidence or tracking origins (provenance) -- AR: We need to think carefully about splitting vs. merging "identical" antibodies. -- We will need to add an "antibody detection assay" to OBI, along with sub-classes.

Action Items:

" vs. "intersection not PRO:000018264"
 * Talk to Darren about representing Cleavage Products -- "lacks_modification PRO:000002971 ! broken protein peptide bond


 * Tim needs to find out if we need to pull out the species for a given uniprot, or will PRO do that for us?


 * Tim needs to ask "how often signal peptides are there, and whether this PRO class PRO:000018264 means 'signal peptide truncated' or not."


 * SCF_PRO_REQUEST:#### -- our own sequential numbers for tracking.  Use ID in OBO in the above form, in submission OBO.  Although this needs a little more thought: it's not just about a broker-unique ID.
 * Alan: the other way is to use a 64-bit GUID, "and then you can choose them randomly without collision."