Issue 5930: rule in the ExperimentDesign package (gene-expression-ftf) Source: Rosetta Biosoftware Business Unit (Mr. Michael D Miller, nobody) Nature: Uncategorized Issue Severity: Summary: The rule in the ExperimentDesign package that only allows a single FactorValue per TopLevelBioAssay hinders proper description of multi-channel experiments and should be dropped Discussion: Raised by Angel Pizarro 3/4/03 "During the MGED Ontology workshop, we came across a situation that does not allow you to describe certain ExperimentDesigns. For instance in a loop design, you compare conditions A,B,C as follows: A -> B B -> C If A,B & C represent different values of the same ExperimentalFactor, how do you encode that a single top-level BioAssay is A -> B? Remember that there is a Rule in the ExperimentDesign package that states that each TopLevelBioAssay can only have a single FactorValue per ExperimentalFactor." Resolution: Reject the Change Revised Text: Actions taken: May 2, 2003: received issue November 6, 2003: closed issue Discussion: Raised by Angel Pizarro 3/4/03: The rule in the ExperimentDesign package that only allows a single FactorValue per TopLevelBioAssay hinders proper description of multi-channel experiments and should be dropped This rule doesn't exist. It was probably in some previous version but is not in the available specification End of Annotations:===== X-Server-Uuid: F7D3E4A3-3C15-41D2-AC5D-A7D3F094E28F From: "Miller, Michael (Rosetta)" To: "'mged mage'" , "'gene-expression-ftf@omg.org'" Subject: RTF Issues Date: Thu, 1 May 2003 12:45:16 -0700 X-Mailer: Internet Mail Service (5.5.2653.19) X-WSS-ID: 12AFA98819003-01-01 All, Finally getting this list of proposed changes out. Many thanks to Ugis for reminding me and helping to bring these out. We are under a bit of a tight deadline due to the lateness of my putting these together, the RTF report is due at the OMG on May 12th. It is exactly the same process as the FTF report was, so I would like to gather thoughts and additional discussions on these issues and any other that people think fall into the scope of MAGE v1.1. Remember, v2.0 is where the major changes will take place, these items are mainly changes to associations to make them more usable and the doc/specifications changes made at the Stanford Programming Jamboree. I will be then mailing out a vote to the OMG Gene Expression Revised Task Force whose members are: Steve Chervitz, Affymetrix; Charles Troup, Agilent Technologies; Ugis Sarkans, European Bioinformatics Institute; Michael Niemi, International Business Machines; Tony Parsons, Pfizer; Michael D Miller, Rosetta Biosoftware Business Unit Michael Michael Miller Senior Application Developer Rosetta Biosoftware michael_miller@rosettabio.com www.rosettabio.com These will change the DTD but looking at it from a developer point of view I don't anticipate much problem being able to cope with either v1.0 or 1.1. Currently I've specifying the doctype as: and will use for documents with the revised changes. I've also been successful at going to the OMG site go fetch the DTD at when receiving the SAX resolveEntity() callback (though it is always good to have a backup location): http://schema.omg.org/lsr/gene_expression/1.0/MAGE-ML.dtd If any changes are accepted, there will then be posted a http://schema.omg.org/lsr/gene_expression/1.1/MAGE-ML.dtd List of Issues for the Gene Expression Revision Task Force (MAGE v1.1) (The numbering is tentative, the OMG will assign their numbers when they are submitted) Michael Miller 05/01/2003 RTF ISSUE 4: The rule in the ExperimentDesign package that only allows a single FactorValue per TopLevelBioAssay hinders proper description of multi-channel experiments and should be dropped Discussion: Raised by Angel Pizarro 3/4/03 "During the MGED Ontology workshop, we came across a situation that does not allow you to describe certain ExperimentDesigns. For instance in a loop design, you compare conditions A,B,C as follows: A -> B B -> C If A,B & C represent different values of the same ExperimentalFactor, how do you encode that a single top-level BioAssay is A -> B? Remember that there is a Rule in the ExperimentDesign package that states that each TopLevelBioAssay can only have a single FactorValue per ExperimentalFactor." Michael Miller replied 3/4/03 "It took me a second but I see the problem and dropping the rule would allow this. But really it seems like the FactorValue isn't associated with the BioAssay now, it is associated with the Channel. When two channels are being used in a traditional Experiment/Baseline way, no one cares much about the FactorValues for the Baseline, just how the FactorValues for the Experiment channel. Logically, I think it might be better to record these two channels as separate BioAssays. They could have been done on separate chips, or n channels could have been hybridized on a single chip, each with a different dosage of a drug, for instance. If that's recorded as one BioAssay, then it would be nice to associate the FactorValues in a formal way to the Channels. But what I'm saying above breaks down if there are ratios or any QTs that use values from the two channels, whether from the channels are hybridized on the same chip or on entirely separate chips. Here my lack of knowledge appears: should the comparison of expression values between two experimental channels fall into higher level analysis or should that be allowed as a transformation? (whether on the same chip or not doesn't seem relevant, it's the same problem) In the first case, then the two channels should be encoded as separate BioAssays. The fact that the same, physical chip was used distinguishes it from performing two, single channel hybridizations on separate chips." Issue: This rule doesn't seem to exist, as far as I can tell. Some one will have to point out exactly where it is in the specification in order for it to be removed, if that is what is decided. --Michael