search

x-atlas-consortia / hra-pop

HRApop
MIT License
0 stars 1 forks source link

Adjust how `similarities` are computed -> leave-one-out #92

Closed andreasbueckle closed 1 month ago

andreasbueckle commented 3 months ago

Idea by @katyb When computing similarities between datasets and AS/extraction sites during HRApop Construction & Enrichment Pipeline, currently (v.10.2), we count the CTs that the dataset contributes to the AS and extraction sites when computing cosine sim between it and the AS/extraction site. This should be omitted. Needs further discussion. Similarity results shown in https://github.com/cns-iu/hra-cell-type-populations-supporting-information/blob/main/validations/violin/validation_violin_plots.ipynb are likely going to get worse:

Untitled

andreasbueckle commented 2 months ago

Issue ongoing. PR #93 did not fix it. Need to revisit at the end of July.