Make info about clusters clearer - different from NBLAST ones

[mmc46]

Info on this difference has now been added to the Features page 37a376e3b8a265c9b305733a9039327bcba625f4

However, most users are not going to read this, when trying the site. What about adding a small question mark icon to the cluster query menu, at the end of 'some part here (clustered by shape)'. This icon would direct people to the right section on the Features page, telling them what the clusters are about.

The other queries are self-explanatory, but I think for this one I think we should provide an easy way to get information.

(Prompted by an NBLAST user contacting the nat-user list)

[dosumis]

Should we be updating clusters to harmonise? Or would that cause other problems?

[mmc46]

The clusters on the NBLAST site exclude intrinsic optic lobe ones, the ones on VFB include all neurons. So I don’t think we can just duplicate them.

(The VFB clusters also need updating, although this is not a high priority currently.)

On 20 Jul 2016, at 16:52, David Osumi-Sutherland notifications@github.com wrote:

Should we be updating clusters to harmonise? Or would that cause other problems?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/VirtualFlyBrain/VFB/issues/958#issuecomment-233993217, or mute the thread https://github.com/notifications/unsubscribe-auth/AByQ07s8CsD7kwU7zjD3CaKZ_IKchpXVks5qXkRZgaJpZM4JQ4_x.

[Robbie1977]

Could we just add descriptions to the cluster definitions as these will start appearing.

[mmc46]

Do you mean descriptions mentioning neuron type? That should be done automatically, by looking at the neuron type of the members. To keep in mind that clusters might include different subtypes.

On 20 Jul 2016, at 17:13, Rob Court notifications@github.com wrote:

Could we just add descriptions to the cluster definitions as these will start appearing.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/VirtualFlyBrain/VFB/issues/958#issuecomment-233999324, or mute the thread https://github.com/notifications/unsubscribe-auth/AByQ02VaTa88213t699XytmiDh8DmEeNks5qXkkwgaJpZM4JQ4_x.

[dosumis]

related #959

[dosumis]

That should be done automatically, by looking at the neuron type of the members. To keep in mind that clusters might include different subtypes.

I was thinking we'd want to replicate the reports currently on the LMB pages. But perhaps we could just have a list of types and then rely on a results page for a list of members. Easy to get a unique list of member types from NEO:

MATCH (i:Individual { short_form : 'VFB_00017796'})-[has_member]->
(:Individual)-[:INSTANCEOF]->(c:Class) RETURN DISTINCT i.label, c

(You'd probably want to filter 'neuron' from the results as too general.)

The major thing missing compared to the LMB pages is blast score (compared to exemplar). But we could potentially move that into Neo - storing it as an attribute on the has_member edge.

If we do that, then it probably makes sense to look at moving storage of clusters into Neo too - dumping from R frame?

[dosumis]

Outline of cluster TermInfo:

1. Need some generic text on how clusters were generated and mapped. @jefferis @mmc46 Can you provide some standard text?

2. Display type of exemplar (if mapped):

MATCH (c:Cluster)<-[:Related {  label: 'exemplar_of'}]-(i:Individual)-[:INSTANCEOF]->
(clz:Class) WHERE not (clz.label ="neuron") RETURN DISTINCT i.label, c.label, clz.label

Queries:

• List members (standard anatomy results page)
• Similar clusters? (Not a first priority, but presumably should be possible given cluster heirarchy @mmc46 @jefferis ?)

[mmc46]

There is info on the cluster pages

http://flybrain.mrc-lmb.cam.ac.uk/vfb/fc/clusterv/3/163/

"Cluster of morphologically similar neurons using affinity propagation clustering.

Score represents an nblast v2 bit similarity score between each neuron and the exemplar. The higher the better. Normalised score (norm_score) is the score for each neuron divided by the self-match score of the exemplar (10728.65). Scores > 0 indicate a match of some sort. Excellent matches will have scores >0.5.

Raw data provided by flycircuit.tw as described in Three-Dimensional Reconstruction of Brain-wide Wiring Networks in Drosophila at Single-Cell Resolution by Ann-Shyn Chiang et al.

Image processing carried out with unu, CMTK, Fiji.

Further processing carried out in R with AnalysisSuite. Clustering used package apcluster. 3D visualisation written from rgl by writeWebGL."

@dosumis Is that enough?

The most similar clusters are already provided on the page, coming from an R object. What would be the best way for you to get this information?

[dosumis]

@mmc46 - text a bit too specific to LMB pages. Note - we don't have any of the actual scores in the KB (as these were never in the LMB mySQL DB). We should probably add them, as well as links to similar clusters, although this then becomes a question of what we should be pulling from the Blast server vs what we store in the KB.

How about

ID: Name: Description: "A Cluster of morphologically similar neurons using affinity propagation clustering based on NBLAST neuron similarity scores (Costa et el., 2016)." [FBrf...]

Clustering version: TBA

Most typical neuron of cluster: -> Insert Exemplar TermInfo Details: Description, Type, relationships.

Thumbnail: As for cluster (as on VFB 1.5) + linkout to LMB (using standard linkout schema).

Sources: Clustering: CostaJefferis Member sources: Dynamically generated. Name, linkout, pub

MATCH (c:Cluster)<-[:Related {  label: 'member_of'}]-(i:Individual)-[:has_source]-(ds:data_source)
...

Licenses: "Refer to individual sources for license details." (????)

Queries: Members of cluster (already in neo) Most Similar neurons (Dependent on pulling info in from R frames or NBlast Server).