Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[go-ontology] Fwd: Reasoner reports and Oort

Paola Roncaglia paola at ebi.ac.uk
Wed Dec 7 08:59:50 PST 2011



-------- Original Message --------
Subject: 	[go-ontology] Reasoner reports and Oort
Date: 	Tue, 22 Nov 2011 10:52:29 -0800
From: 	Chris Mungall <cjmungall at lbl.gov>
To: 	GO Ontology Editors <go-ontology at mailman.stanford.edu>



Over the years I've produced various ad-hoc reasoner reports for the ontology group, mostly David and Tanya. These are particularly necessary for the time leading up to the incorporation of new logical definitions, where we have to go back and retrospectively fit the asserted links in the ontology such that they match what a reasoner would produce.

These are being consolidated into a single report that gets generated by Oort, powered by standard OWL reasoners such as ELK behind the scenes. Oort is easy to install, so the goal is to take me out the loop here!

In fact the reports are a temporary measure - in the long run we should not be report driven, everything should be done at the time of ontology editing. We'll discuss this in January - for now the reports are useful.

I'm attaching a copy of the report for the current ontology plus all of bp_xp_cc added.

** don't start working through the report just yet!! **

The idea is just to get a feel for the current state of the ontology, plus a chance to give feedback on the structure of the report. The slew of SF items I submitted last week came from some of these.

The other important thing to bear in mind is that in future *the published ontology will be the post-reasoned one*, so you should have confidence that the inferences are correct (or rather the input axioms are correct).

* "EXISTS, ENTAILED"

This means the link is asserted in the ontology, but we could infer it from other axioms if we wanted. No action is required here. Eventually we will move to an approach of calculating these dynamically. As you can see, most of the report falls into this category. Eventually we can calculate these dynamically.

* "EXISTS, NOT-ENTAILED"

The link is asserted in the ontology, and it's between two defined classes, yet the reasoner could not recapitulate the link. D&T know this as the "abduced link report". Often this is a sign something is not quite right somewhere.
There's a lot of regulation terms in there!

e.g.
	EXISTS, NOT-ENTAILED     SubClassOf(GO:0010611 "regulation of cardiac muscle hypertrophy" GO:0043502 "regulation of muscle adaptation")

The subclass link between the regulation classes has been asserted by one of you, and this _would_ be entailed _if_ there was the appropriate relationship between CMH and muscle adaptation, which there is not.

this report may be a little oversensitive. there is a subtle issue to do with the "X part" terms. Also you can see the problems the "cellular X" terms cause here when it isn't applied consistently.

Note that with the eventual goal of dynamic inferencing, this report will be obsolete as we will only assert minimal links

* "NEW, INFERRED"

Straightforward new links, i.e. same as what you would get with the AIL panel in OE.

* "NEW, TRANSLATED"

you can ignore this for now. Cases where we have the logical definition but not the trivial all-some that follows from it.

* "REDUNDANT"

Same as the redundant links visible in OE. Harmless, and in fact these are auto-removed by Oort for the release version.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.stanford.edu/pipermail/go-ontology/attachments/20111207/4adb6610/attachment-0001.html>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: go-reasoner-report.txt
URL: <http://mailman.stanford.edu/pipermail/go-ontology/attachments/20111207/4adb6610/attachment-0001.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Attached Message Part
URL: <http://mailman.stanford.edu/pipermail/go-ontology/attachments/20111207/4adb6610/attachment-0002.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Attached Message Part
URL: <http://mailman.stanford.edu/pipermail/go-ontology/attachments/20111207/4adb6610/attachment-0003.ksh>


More information about the go-ontology mailing list