mellybelly
commented
10 years ago Another idea here is to have a total computation for all models against the patient, so that if, for example, the fish recaptulated some aspects that the mouse complemented, you'd then get an overall score for all of the models developed for that patient.
as came up on the UDP call today, is the idea of identifying how well a model recapitulates a portion of phenotypes in a human disease.
for example, we want to ask how well a fish's skeletal phenotypes match that of a patient.
right now, we only see an overall similarity score, which heavily penalizes phenotypes that are not held in common. but we might want to isolate a subset of the phenotypes to compute a "local" similarity score just for phenotypes of interest.
one way we can accomplish this is to use the subset of both the query and target phenotypes, and then compute the similarity scores between them. in an interface, ideally you'd want to specify what subset of the phenotypes the user wants to consider. these could be specified either individually, or as subclasses of a more general class (like abnormality of the skeletal system).
this might extend the compare owlsim call, and other services described in #349 & #403