Cache fix in mondo_score_utils, now faster #27
Conversation
Got rid of profiling tool in runner, forgot to this before
|
|
||
| descendants_list = mondo.descendants([prediction], predicates=[IS_A], reflexive=True) | ||
| for mondo_descendant in descendants_list: | ||
| if ground_truth in omim_mappings(mondo_descendant): | ||
| if ground_truth in omim_mappings(mondo_descendant, mondo): | ||
| # prediction is a MONDO that maps to a correct OMIM via a descendant | ||
| return PARTIAL_SCORE |
There was a problem hiding this comment.
It seems that we should provide full score (full rank) if a disease is in the same phenotypic series, but not necessarily if it is a Mondo descendent. I am not sure if that is what the code above is doing, but if so it would be good to add a comment and an example of what does and does not count (lines 72-74). I am not sure that we want to return a partial score for anything else, and it would really be better just to report ranks
There was a problem hiding this comment.
Looking at the picture and description at #14 I assumed that all mondo descendants of a mondo term corresponding to a pheontypic series (I mean all descendants of the parent or "grouping" term, for lack of better words -- the one which does not have an OMIM ID on its own) are leaf terms. If this is not the case, @pnrobinson we need to discuss, I am not finding much info about phenotypic series online ...
Also, do note that PARTIAL_SCORE is there in case we were to decide that we want to assign partial scores, which for now we are not doing (both FULL_SCORE and PARTIAL_SCORE are mapped to 1).
…er in one function in the docstring
|
Fixed CI error, it was a docstring test failing, |
|
Merged since no more conflicts were present, some of the mentioned things that still need to be done are now issues and can be done independently after the merge. |
No description provided.