TermHub: Neighborhood analysis

jhu-bids / TermHub

Web app and CLI tools for working with biomedical terminologies. https://github.com/orgs/jhu-bids/projects/9/views/7

https://bit.ly/termhub

GNU General Public License v3.0

11 stars 10 forks source link

TermHub: Neighborhood analysis #84

Open joeflack4 opened 2 years ago

joeflack4 commented 2 years ago

Tasks

[ ] 1. Data visualization: nodes = concept sets; edges = connections to other concept sets that share 1+ concepts.
[ ] 1.2. Edge weighting: Weight is inversely proportional to either (a) the size of the overlap, (b) the jaccard measure (|intersection|/|union|), or (c) something else (perhaps selectable). This can be visualized by either (a) visual length of edge, (b) visual width / boldness of edge, (c) display text showing % overlap.
[ ] 1.3. Interactivity: Users could click on an edge to get an overlap report (# codes, # patients, # records).
[ ] 1.4 Interactivity: Lasso a set of concept sets to get a report [not yet specified] about the concept sets in the lariat.

hlehmann17 commented 2 years ago

Ideally, there would be a graph of concept sets (nodes), connected by arcs, where the length of an arc is inversely proportional to either the size of the overlap or the jaccard measure (|intersection|/|union|) or something else (perhaps selectable). Users could click on an arc to get an overlap report (# codes, # patients, # records), or lasso a set of concept sets to get a report [not yet specified] about the concept sets in the lariat.

joeflack4 commented 2 years ago

@hlehmann17 Thanks, that's really good. I broke the tasks down into different steps. I think that we could get up to 1.2 done a lot faster. 1.3 probably not bad too. 1.4 I haven't seen in a JavaScript library of this type before, so I'm guessing this would potentially take a lot more time.

Just some minor semantics; I used the word edge instead of arc. I usually say edge. I looked up the difference, and it looks like arc is typically used when the edge is directed, which I don't think is the case here.

For 1.2, I think it would be nice to have (c) a popup. I think (b) edge width, would be better than (a) edge length. I could imagine the graph could look really wonky with edge length unless we added some downscaling, but we'll see whenever we get to working on this.