search

satijalab / azimuth

A Shiny web app for mapping datasets using Seurat v4
https://satijalab.org/azimuth
GNU General Public License v3.0
108 stars 31 forks source link

Azimuth and integrated data #210

Closed tirk999 closed 5 months ago

tirk999 commented 7 months ago

Hello everyone!

I want to use Azimuth to annotate two datasets that I have, by using a modified RunAzimuth() function (I had to adapt it to small datasets). Each dataset corresponds to a different replicate of the same experiment. As far as I know, Azimuth works with the "counts" slot of the RNA assay. Therefore, I assume that I cannot use the "integrated" assay to use Azimuth, in case I have one. Anyway, I am working with Harmony, which does not create an integrated assay. When I try to run my function with the integrated assay obtained by Seurat integration, I indeed obtain this error. Error in h(simpleError(msg, call)) : error in evaluating the argument 'x' in selecting a method for function 'colSums': no 'dimnames' attribute for array My first question is:

  1. Is it correct to always use the RNA assay when I run Azimuth?

When coming to wanting to label both datasets,

  1. Should I run Azimuth on each dataset separately or
  2. Should I merge the data (not integration) and run Azimuth on the "RNA" assay of the merged data?
  3. Moreover, I understand that Azimuth applies SCT so I suppose there is no need for LogNorm or SCT normalization before running Azimuth. Is this correct? Thank you a lot.
Gesmira commented 5 months ago
  1. Yes, you should run Azimuth on the RNA assay with unnormalized data, as the app will normalize the data. 2 and 3. This is up to you, we don't have a clear recommendation for this, but it should give similar results either way.
  2. Correct, you do not need to normalize your data before running Azimuth.