ITGAE and RORA point to granzyme positive t resident memory populations, and clusters of these tend to be more unique to certain sections of the gut, driving down the number of datasets they are found in (dataset diversity metric)
Cell annotations based on two rounds of subclustering with scVI integration, sample_id as batch key. Cell annotations were refined by expert feedback based on "prelim annotation" on CAP. Prelim annotation is renamed "hgca_celltype_v0" here and the feedback on the "prelim annotation" is available by selecting a previous version of the Integrated Human Gut Atlas on CAP.