[go] AmiGO and annotation qualifiers

[Judith Blake]

hummm
I think the case that Harold was saying, and that we currently have in 
other annotations here at MGI, is that we have

x A
x NOT A

both lines of evidence exist at this point.

In some cases, different experiments give different results

In the case that Harold discussed, the issue really was that we don't 
properly distinguish isoforms, so from the gene level, the two are 
combined whereas if you could represent each isoform, the one would be 
x-1 A and one would be x-2 NOT A.

judy

Chris Mungall wrote:
>
> On Jan 31, 2008, at 2:04 PM, Jim Hu wrote:
>
>>
>> On Jan 31, 2008, at 1:30 PM, Chris Mungall wrote:
>>
>>>
>>> Unambiguous is good.
>>>
>>> A related question:
>>>
>>> Let's assume there is a NOT annotation for GO:0008047 enzyme 
>>> activator activity (is_a parent of proteasome activator activity)
>>>
>>> If you are on the page for GO:0008538 proteasome activator activity, 
>>> would you expect to see this annotation?
>>
>> I'd expect to see this iff there is an unqualified annotation to the 
>> same product. In other words, this is only useful as a conflict flag 
>> IMO.
>
> So let's summarise
>
> Given an is_a hierarchy:
>
> A
>   B
>     C
>
> And annotations of gene product x, y and z - here, numbered for 
> reference purposes:
>
> [1] x NOT A
> [2] x B
> [3] y B
> [4] y NOT C
> [5] z A
> [6] z C
>
> The same information shown in the DAG:
>
> A    : NOT-x, z
>   B    : x, y
>     C    : NOT-y, z
>
> We have a contradicting pair of annotations:
>
> [1]+[2]
>
> the other pairs are non-contradicting:
>
> By your criteria the page for B should show any annotations of gene 
> products that are purported to have property B, *plus* any annotations 
> that contradict the annotations shown. i.e. the page would show:
>
> [1] x NOT A
> [2] x B
> [3] y B
> [4] y NOT c
> [6] z C
>
> (structured in some way that reflects the DAG and with the 
> contradiction visually highlighted)
>
> Note that [5] is not shown because A is less specific than B, and it 
> contradicts nothing on the page
>
> Note that the way AmiGO (and I imagine most annotation browsers) are 
> set up now is such that the following are shown on the detail page for B:
>
> [2] x B
> [3] y B
> [4] y NOT c
> [6] z C
>
> The fact that [2] has a contradiction with [1] becomes apparent only 
> as focus is shifted up the DAG. We have a simple algorithm that goes 
> down the DAG (or propagates genes up depending how you look at it), 
> ignoring the NOT reversal rule.
>
> Which do people prefer?
>
> Either way AmiGO will most likely keep the current algorithm for the 
> next release.
>
>>>
>>>
>>> I think most people's answer would be no, this would be too confusing.
>>>
>>> However, it is the case that anything that is not an enzyme 
>>> activator is by definition not a proteasome activator activity.
>>>
>>>
>>>
>>> On Jan 31, 2008, at 11:21 AM, Michael Ashburner wrote:
>>>
>>>> I prefer demo3 - it is clear and unambiguous.
>>>>
>>>> Michael
>>>>
>>>> Michael
>>>>
>>>> On 29 Jan 2008, at 14:00, Amelia Ireland wrote:
>>>>
>>>>> Hi GO Consortium,
>>>>>
>>>>> There has been some discussion in the web presence working group 
>>>>> about the display of NOT annotations and annotations with 
>>>>> qualifiers. There are two issues; how to display the annotations 
>>>>> with qualifiers, and how gene product counts for terms with 
>>>>> qualifiers should be displayed. We'd like to get some feedback 
>>>>> from the consortium on these issues before releasing the new AmiGO.
>>>>>
>>>>> Annotations With Qualifiers
>>>>> ===========================
>>>>>
>>>>> I've made some mock ups of different possible arrangements here:
>>>>>
>>>>> http://www.ebi.ac.uk/~aji/demo.html
>>>>> http://www.ebi.ac.uk/~aji/demo2.html
>>>>> http://www.ebi.ac.uk/~aji/demo3.html
>>>>>
>>>>> These pages are viewing the GPs annotated to a term (and its 
>>>>> children), but a similar arrangement will be used to view the 
>>>>> terms with which a GP is associated.
>>>>>
>>>>> Which arrangement is the clearest? Things to consider:
>>>>>
>>>>> - how easy is it to see inconsistent annotations, e.g. the same GP 
>>>>> has a normal and a NOT annotation?
>>>>>
>>>>> - is it clear what the qualifier applies to?
>>>>>
>>>>> - some views (e.g. having the associations with 
>>>>> operators/qualifiers separated out) require extra calculations to 
>>>>> generate and will hence be slower
>>>>>
>>>>> - should the default view contain NOT annotations? Is a GO newbie 
>>>>> going to understand them?
>>>>>
>>>>> - is the 'nn gene products' link unambiguous in meaning? Is it 
>>>>> helpful to know how many GPs there are annotated to a term?
>>>>>
>>>>>
>>>>> Gene Product Counts
>>>>> ===================
>>>>>
>>>>> The WPWG discussed generating separate totals for numbers of GPs 
>>>>> annotated to a term without any qualifier and the no. of GPs 
>>>>> annotated to a term with each of the other qualifiers (currently 
>>>>> the GP count for a term *includes* any GPs annotated with 
>>>>> 'contributes_to' and 'colocalizes_with', but *excludes* NOT 
>>>>> annotations). How useful would this be, and is it intuitive to 
>>>>> split the counts up thus, or is it confusing (especially for GO 
>>>>> newbies)? There is also this to bear in mind:
>>>>>
>>>>> On 24 Jan 2008, at 20:11, Chris Mungall wrote:
>>>>>> On Jan 24, 2008, at 11:08 AM, Seth Carbon wrote:
>>>>>>> *) ask chris about splitting the gp count by qualifier (seth)
>>>>>>
>>>>>> you mean by contributes_to, the NOT operator etc?
>>>>>>
>>>>>> unfortunately it is expensive to pre-compute the gp counts in any 
>>>>>> way that a single gp can end up in two partitions.
>>>>>>
>>>>>> we currently partition by gp db name - no gp can be in > 1. this 
>>>>>> has the advantage that we can get the correct total by summing 
>>>>>> the numbers at query time.
>>>>>>
>>>>>> however, if we have a partition that does not split gps in this 
>>>>>> way we can't sum across the partitions, which means we have to 
>>>>>> pre-compute for all combinations.
>>>>>>
>>>>>> This may not be so bad for the qualifiers however, since the 
>>>>>> qualifiers are relatively rare.
>>>>>>
>>>>>> would you really want this for the NOT operator though? remember 
>>>>>> the semantics are easy to get confused here. Would you want:
>>>>>>
>>>>>> all gps that are NOT GPCRs (ie include those that are asserted 
>>>>>> not-TM receptor in the count) - negation propagates up the graph
>>>>>>
>>>>>> all gps that have a NOT annotation to some kind of GPCR - 
>>>>>> negation propagates as normal
>>>>>
>>>>>
>>>>>
>>>>> Any feedback would be gratefully received. Thanks!
>>>>>
>>>>> Amelia / the Web Presence Working Group.
>>>>>
>>>>> -- 
>>>>> Amelia Ireland
>>>>> GO Editorial Office,
>>>>> European Bioinformatics Institute, UK.
>>>>> Carbon neutral driving: http://www.targetneutral.com/TONIC/index.jsp
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>> =====================================
>> Jim Hu
>> Associate Professor
>> Dept. of Biochemistry and Biophysics
>> 2128 TAMU
>> Texas A&M Univ.
>> College Station, TX 77843-2128
>> 979-862-4054
>>
>>
>