Closed miltargid closed 2 years ago
@miltargid, you can save 'harmony-corrected pca' in the 'X_pca' slot, or alternatively, do over clustering yourself, such as sc.pp.neighbors(adata, use_rep = 'X_pca_harmony')
and sc.tl.leiden(adata, key_added='over_clustering', resolution = 15/20/30/..)
, and then do celltypist.annotate(adata, model='xxx', majority_voting = True, over_clustering = 'over_clustering')
Hi, Thank you for providing the tool! I integrate the data with harmony which 'corrects' the PCA embeddings as a means of batch correction. This corrected PCA is then used for downstream clustering. If I were to use an integrated batch corrected data, would you suggest to save 'harmony-corrected pca' in the 'X_pca' slot of the adata object, since I see that in some modes the classifier uses PCA embeddings?
Kindly advise, Thanks and Kind regards,