User confusion in the neighborhood view; make note of indirect and direct relations

[ValWood]

Whilst

PomBase were looking at this (mainly to find a way to display neighbouring terms in Canto), we thought this looked a little odd.

[ValWood]

We were wondering why not display as a simpler list. Currently the term is in the heading "Parents of detection of stimulus...." and than repeated on every row....

It might be better to make the therm name more prominent at the top, and then not repeat.

Also the labels 'subject' and 'object' are a bit strange in this context.

Finally I'm not sure now exactly what this is meant to show. At first I thought this was only the 'immediate' parents, but for this term this doesn't appear to be the case. Will demonstrate....

[ValWood]

The neighbourhood terms are highlighted, what doe this mean?

[cmungall]

Not following the connection between these.

• for the first, I agree the table may be overkill and a simple list may be better
• no idea why the quickgo screenshot has things colored. how did you get here?

[ValWood]

I added the terms in the "neighbourhood" to the basket and viewed them in the ancestor chart.

I can't figure out how you decide what is "neighbourhood'. For example why 'detection of stimulus' and 'mitotic cell cycle checkpoint' they aren't direct ancestors....

[kltm]

@ValWood could you give the URL for the AmiGO instance that you're using, just so we're sure what version that we're looking at? Since the default image that we present in [Graph Views] comes from QuickGO, we're not always sure the exact methods they use to generate it. The internal AmiGO graph builder is also available there, but it must be selected separately (it's a link at the bottom of the [Graph Views] tab section). Also, what is the "cart" that you're referring to in this case? AmiGO did at one point have have some cart functionality, mostly for use with Galaxy, but that should mostly be removed from common interfaces. If you could give an example of its use...

[ValWood]

I thought we were using the live version, as this is what I have bookmarked http://amigo.geneontology.org/amigo/term/GO:0072442

However, the "neighbourhood" tab does not appear to be present in the live version? Maybe I took this screenshot from a screen share..... @mah11 @kimrutherford which AmiGO were we using? Can you remember?

(Seth, I don't have a problem with the QuickGO view/image. I just used QuickGO to highlight the terms which were displayed under 'neighbourhood' so I could explain that I did not understand what constituted "neighbourhood".)

[kltm]

Possible places to have seen the [Neighborhood] tab are:

• http://tomodachi.berkeleybop.org/amigo/term/GO:0072442 (labs version to go live soon)
• http://ec2-23-22-182-58.compute-1.amazonaws.com:9999/amigo/term/GO:0072442 (your fypo instance)

A bit of the discussion around the "full" neighborhood (all edges) display can be found here #220.

[ValWood]

OK so Neighbourhood is new. What determines why the terms which are not close ancestors are displayed? I still can't figure this out....

[kltm]

The use case for [Neighborhood] is that it would contain all direct links between two terms, not just the main ones that are typically used (something desired by the Planteome people). To be explicit and help see what might be going on, could you provide:

• the term IDs for the two terms that are under consideration
• the full URL for the page where you're seeing them in the [Neighborhood] tab

[cmungall]

Yes, this isn't just useful for planteome. In GO, it's useful to see direct links over relationships that would not be shown in the standard graph or hierarchical views. This includes for example has-part, process to chemical or cell links, etc.

The full discussion is here: #220

[ValWood]

OK, so it should display ALL direct links. When you say "all direct links" do you mean all the terms it is related to, or just the immediate ancestors? The screen shot above (which I still don't remember where it was from), does not display either of these, it displays the 5 terms I highlighted in the QuickGO graph. (I'm not disputing its usefulness. I'm just trying to understand what it is displaying).

[cmungall]

Slightly dodging the question here, but part of the issue is that there are conflicting definitions of 'related to' and 'immediate ancestors'. My action item is to clarify these and document the behavior.

It may be current behavior is correct by some definition but suboptimal from a user perspective. For example, 'detection of stimulus involved in X' is always directly "connected" to 'detection of stimulus' by an axiom. But what the user wants to see is direct parents in the relaxed inferred graph

[ValWood]

OK when you have documented can you let me know and I'll see if I understand this then? At present I'm lost.....

[kltm]

@cmungall Has there been any activity on this?

[cmungall]

perhaps we can close this and open distinct tickets.

The apparent redundancy issue is addressed here: https://github.com/geneontology/amigo/issues/366

[kltm]

On advice of @cmungall, reopening to tackle what ever happened to the indirect/direct notice in the neighborhood view.

[kltm]

Well, obvious answer: the neighborhood view is (now) just parents and children, so no need for further marking. @cmungall Can you find an example where this is not true?

[kltm]

More talk with @cmungall , may actually be a difference here if the fixes did not make it up back up far enough: getLineageShuntGraphJSON vs. (getNeighborsJSON and getNeighborsLimitedJSON)

[kltm]

Which...ties it back to: https://github.com/geneontology/amigo/issues/220, so we /want/ to see these relations, yes? So that brings us back to allowing these (for the "Planteome" use case), but marking them.

[ValWood]

Looking through tickets I opened (just because I landed in the AmiGIO tracker tonight). Is this still relevant? It does not bother me much because I never use or demonstrate this view....