Support search/sort in relations tables and color concept nodes, by UMLS semantic type

[b]

This user story has two parts (Will need to bring in UMLS database).

1) Use the UMLS semantic type system to allow the user to filter the triple table. For example, show only triples containing a node of type 'chemicals and drugs'.

2) Use this information to color the nodes in the concept map network.

[Richard Bruskiewich]

Farzin will tackle part 1). We don't need the full UMLS database ... we have the code details in a separate file which she can use to add a drop down menu to the relation result tables for searching purposes (based on semantic type). The search refinement needs to be made to the back end as well (probably by RB)

[b]

It might be useful to map the UMLS semantic types (of which there are about 50) up to the smaller, more general set of 'semantic groups'. The mappings are available here: http://metamap.nlm.nih.gov/SemanticTypesAndGroups.shtml

[Richard Bruskiewich]

We've now rolled out a first iteration of this issue, which provides a simple drop-down menu to pick from four options: "No Filtering" (Default), "Disease or Syndrome", "Genes or Genomes" and "Pharmacologic Substance".

We are working on a 2nd iteration which will provide a popup window list of semantic types to select from.

[Richard Bruskiewich]

Item 2) is a cytoscape.js node coloring issue (assigned to Kenneth)

[b]

@KennyBruse How far away are we from node coloring in the graph? If this can be done in a day or two, lets get it in. If not, resolve this issue and we will keep this in mind for V2.

[Richard Bruskiewich]

I've just been taking a closer look at item 2) node coloring. I've concluded that it is non-trivial in that nodes only have concept_semtypes in SemMedDb in the context of relations (predications) within which they appear, and that individual concepts may have several distinct semtypes to only apply to one subset of predications or another.

I'll see if I have any feasible way to figure out the context of the node for coloring. No guarantees (also assuming only genes, diseases and pharma nodes are targets at this time...).

The best first step we can take here might be to simply to color "ENTREZ" versus "META" concepts but since most genes are gene symbol names, their status as a gene is betrayed by their concept name.

[b]

http://semanticnetwork.nlm.nih.gov/SemGroups/ would be my take on this. This issue is low priority though. Its a nice-to-have but not vital for this release.

[Richard Bruskiewich]

Functional node coloring by major groups GENE, DISOrder and CHEMistry (including pharmaceutical compounds) is now working, including persistence across save/load (concept) map files created from this point onward.