search

monarch-initiative / helpdesk

The Monarch Initiative Helpdesk
BSD 3-Clause "New" or "Revised" License
7 stars 0 forks source link

Pre-computed phenotype profile similarities #34

Closed eric-czech closed 2 years ago

eric-czech commented 2 years ago

Hi,

Have you ever released pre-computed similarity matrices from the phenotype analyzer when the profiles are seeded based on genes or diseases? I.e. I'm looking for an N disease x N disease matrix with all the current API results available at https://api.monarchinitiative.org/api/sim/search for a single request.

I would like to do this in bulk without the API if possible and thought I'd ask for suggestions.

Thank you!

nlharris commented 2 years ago

Hi @eric-czech, thanks for submitting this question. Hopefully @putmantime or @kshefchek can help you.

kshefchek commented 2 years ago

I have generated disease x disease distance matrices in the past, but we have not considered making them a part of our data releases. There are a number of ways to define distance between diseases, in both the distance metric and how the annotations are selected (if using an annotation based method). We've also created pipelines to generate graph embeddings of our ontologies, and presumably this could also be used to generate a distance matrix - @justaddcoffee is leading that effort. @eric-czech if you have an idea of what you would like for your analysis we could point you in the right direction.

eric-czech commented 2 years ago

Thanks @kshefchek! The phenodigm distance metric and HPO terms chosen based on diseases would be great (if that's what you mean by annotations). Graph embeddings for diseases would also be great -- are those somewhere public?

kshefchek commented 2 years ago

@eric-czech it might be easier to discuss over a quick call, could you reach out to me via kent@tislab.org ?

nlharris commented 2 years ago

@eric-czech did you get what you needed from Kent? Can we close this ticket?

eric-czech commented 2 years ago

Yep, all set! @kshefchek gave us an intro to @pnrobinson, who we spoke to today. We haven't figured out anything yet but we're working on it, so there's no need for this issue anymore.