June 1st 2011, 12PM EST
–Seth :
leave the phenotype module for later
–Yuri:
2 new prop tables ( in the genotype module) environmentprop phenstatementprop
–Naama:
the question is how do we use the property tables. What should go in there, and do we want to structure the values by attaching cvterm FKs to it.
–Sook:
terms can be post-composed using cvterm_relationship
–Lukas:
postcomposing terms does not include description of the experiment
–Seth:
there is no justification for adding cvalue_id if there is another way in Chado to post-compose terms. the values in these 2 prop tables could be stored elsewhere in the ND schema (these are props of the experiment). Since nobody in this group, besides Yuri and Maren is using environmentprop and phenstatementprop, this is not the right group for deciding whether we add these or not.
Decision
Need to send an email to gmod-schema regarding adding the 2 vanilla prop tables (Yuri will do this).
–Yuri:
nobody has replied to the previous email about these 2 tables
–Scott:
people don’t use/ask for these 2 prop tables because they don’t need it, or they are not aware this is an option. I’m inclined to do nothing for now, but Yuri should send the email because there are now more Chado users.
–Yuri:
nd_experiment_protocolprop Add nd_experiment_phenotypeprop
property of the linking table - applies both to experiment and the protocol. same for the phenotype .
–Seth:
nd_experiment_protocol :
–Yuri:
this allows flexibility in the schema
–Seth:
Can these be 1 protocol with 2 different experimentprops? if you are testing the reaction of your plant to a substance, this is part of the nd_experiment (2 separate props within nd_experiment) and the protocol remains the same.
The only place with a prop table for a linking table is nd_experiment_stockprop - it was important to provide information on that relationship - how the link between the stock and the experiment was made. This is not the case with experiment and protocol. The stock-experiment value cannot be put anywhere else, because the stock links to multiple experiments and an experiment can have multiple stocks.
–Yuri:
I can see somebody wanting to be able to provide the context for these 2 objects but currently nobody does that. I don’t have a problem with adding prop tables because I don’t think it clutters the schema (while Naama and Sook think it does)
–Seth:
adding prop tables to linking tables changes the schema by saying we can add context to the relationship
–Yuri:
there’s also the question of consistency of the prop tables in the ND module. If we want that, I’m fine with adding cvalue_id to all the prop tables in ND, or at least to the ones I find useful. It’s an optional field. I find it very useful.
–Naama:
we have here 2 fundamental issues: 1. how do we break the value-phenotype connection in the phenotype module 2. how do we use the prop tables? (and cvalue_id, is this needed?)
–Yuri:
looked at postcomposition in the phenotype_cvterm table - we need the value field for some of the data. e.g. we have the date sometimes as part of the description. ‘Stem diameter at ‘ + a certain date.
–Naama:
Do we change the schema because of the data or do we change the data so it would fit into the schema?
–Seth:
it’s a valid argument. We’ve run into this last week talking about postcomposing terms.
Need to solve first the problem with post-composing terms . Solve the cases when you cannot pre-compose. This is a general problem , and should be looked into in the cvterm module (cvterm_relationship or somewhere else) because this is not a specific problem in the phenotype or the ND module.
–Yuri:
If we use cvterm_relationship we’ll need a way to link my object to a cvterm_relationship table . sometimes I just want to add a unit , and this seems like making things more complex.
–Seth:
you can just link to the unanimous term. This is the same as adding a new cvterm and linking to it.
–Naama:
would be good to email the gmod-schema list and see if anyone is using cvterm_relationship for post composition.
–Scott:
I can imagine this as a Tripal module that can help others.
–Seth:
thinks Phenote allows post-composition of terms (need to check with ZFin)
–Scott:
Zfin does not use Chado, but they do use the cvterm module. Maybe Nicole Washington knows, if she still works on Phenote. I’ll email her and find out.
–Naama:
thinks next step should be going back to the gmod-schema list asking about post-composition of terms
–Seth:
and CC Zfin Flybase helpdesks - most obvious people who might be post-composing terms. Will send the email.
Decisions:
1. decision on post-composing terms . How?
2. revisit the phenotype module for breaking the value-phenotype ties
3. better definition of how we use the prop tables