Open nlwashington opened 9 years ago
Not sure what you mean - what would a typically column and row heading + cell value be?
columns and rows would be each category: Gene | Variant(s) | Disease | Phenotype | Model | Pathway | Drug | Homolog
the value of each cell would be count of association statements for each paired category. For example, disease-gene associations would fill in the cell of Disease (col) x Gene (row), and Gene-Gene interaction associations would fill in the Gene x Gene cell, asserted models of disease would be Disease x Model, etc. My thought is that this would just be the counts of the result of the cypher queries that are going into Golr. These would be the direct assertions.
It's mostly just a survey of the system that we get for free with storing the associations in Golr.
let's discuss on Thursday's call...
@kltm has something for that https://github.com/geneontology/amigo/blob/master/javascript/web/FacetMatrix.js
Just call this with the following set:
We don't have quite enough populated yet, it would be a boring 3x1 table at the moment: /solr/golr/select/?q=:&rows=0&facet=true&facet.field=object_category&facet.field=subject_category
Functionality is very well hidden in amigo unfortunately... and seems to be switched off in production http://amigo2.berkeleybop.org/amigo/search/annotation Click the 'matrix' button
Another example / possible behavior test, Bardet Beidl
This disease has several subtypes. Each subtype has its own unique features (phenotypes) and shared features. A researcher wants to see the landscape of phenotypes for each. At the moment they can go to
http://beta.monarchinitiative.org/disease/DOID:1935 or http://beta.monarchinitiative.org/labs/widget-scratch/disease/DOID:1935
In the latter they can drill down to subtypes in the facet, but no landscape.
they follow a link to a 'landscape view', which is the matrix with the following settings:
Here they see that one phenotype 'Retinal dystrophy' is shared, bit otehrs are unique
Extensions: we would also like to see inferred overlap, here we need some indication of relationships, but this obviously bleeds into a PhenoGrid use case
interesting. let's talk about this in a ui call.
Actually for this particular use case, it would be better to just use the phenogrid widget directly, but restrict the search to be across all subtypes. E.g. currently on http://beta.monarchinitiative.org/disease/DOID:1935 in the 'compare' tab, the various other subtypes of BB are not shown; this is as expected because other diseases are better matches (reflecting the diversity of BB). So what we want here is to see this with all BB subtypes as the first N columns
We will be able to facet on the source of associations once https://github.com/monarch-initiative/monarch-app/issues/809 is satisfied.
Now that https://github.com/monarch-initiative/monarch-app/issues/809 is closed; wondering if we can revive the brainstorming on this.
once we have all the golr documents set up, we should be able to generate a nice non-directional pairwise association matrix overview.
basically i want a 2D table that shows the counts of pairwise associations that live in our system for: Gene, Variant(s), Disease, Phenotype, Model, Pathway, Drug, Homolog
this can be a nice table that lives on our landing page.
@ccondit should update this ticket with the golr document-category tags to use. @kshefchek can work on creating the UI component for the table (in labs?).
sugar on top would be to have facets. for example, by taxon.