Eli Roberson (he/him)Adventures in #GeoMx data analysis. The available tools are fine for loading info from a nanostring software processed DSP file. They suggest some segment (tissue piece) QC based on the area of the slide and number of trimmed reads, etc. Also fine.
Then I pulled the actual count data. I plotted the total number of UMIs (x) versus genes detected (y), defining detection as at least 10 UMIs in a sample. In these samples (fixed eye), I don't see a reasonable relationship between UMIs and detected gene count until a minimum of 50,000 reads. Somewhat linear relationship that saturates at >200,000 total UMIs. Based on that, I dropped samples with fewer than 50k UMIs since they aren't likely to be very useful.