Indicate number of unmatched phenotypes per column

monarch-initiative / phenogrid

The phenogrid widget

13 stars 14 forks source link

Indicate number of unmatched phenotypes per column #245

Open jmcmurry opened 8 years ago

jmcmurry commented 8 years ago

This is really more of an owlsim / scigraph issue but manifesting here.

It is not clear why we're seeing higher scores for diseases that have fewer phenotype matches. I expected this to be the result of slightly different scoring, or weighting by IC score, but this appears not to be the case?

@kshefchek @jnguyenx ?

jnguyenx commented 8 years ago

Scoring is done by the owlsim server. @cmungall must have answers for that.

cmungall commented 8 years ago

we have an option to show unmatched phenotypes for the query, but not for the matches. If we did, you'd see the lower scoring ones would have many unmatched phenotypes

jmcmurry commented 8 years ago

Ah right; of course. Perhaps then, phenogrid could at least have a number for each column showing how many unmatched phenotypes there are? That would help make more sense of reports like this.

mellybelly commented 8 years ago

I've always disliked the hidden unmatched terms. If we implement a hierarchical y axis there might be a way to render them differently rather than hiding them

jmcmurry commented 8 years ago

On a related note, when we get to phenogrid from a disease page (compare tab) I can't figure out why it is that the top disease hit isn't either:

A) a different disease altogether OR B) the same disease as the query disease but scoring a 100% profile match

We have neither A nor B now: the top hit is always identical in name/ID to the disease that spawned the query; however the match scores can be well below 100; for example https://monarchinitiative.org/disease/OMIM%3A127750#compare has a match score of 79. What am I missing?

cmungall commented 8 years ago

In this case, what is in owlsim doesn't match what is in solr. This is odd as the owlsim data is dumped directly from golr. It could be a data synchrony issue, or it could be something wrong with the golr-exporter script. @jnguyenx and @kltm can explore more tomorrow.

cmungall commented 8 years ago

Checked on beta, same issue. But it also showed odd fly matches.

Following up, we should not be including inferred phenotypes like these https://beta.monarchinitiative.org/gene/FlyBase:FBgn0040074

(this belongs in a separate ticket, just noting here for now)

EDIT: not a phenogrid issue, now tracked here: https://github.com/monarch-initiative/monarch-app/issues/1243

yuanzhou commented 8 years ago

Let me know if there's any changes better to be made in Phenogrid, or just help with testing.

harryhoch commented 8 years ago

@yuanzhou, don't worry about this for now.