Azimuth prediction and mapping scores - obtaining scores for all possible cell types

[HarrietRB]

Hello! I was wondering whether you can help.

I have been annotating my data and have obtained the predicted cell type, prediction score and mapping score for each of my cells. There are a total of 19 cell types present in my reference dataset.

All of my cells have received very high mapping scores, but there are many (~50%) that have low prediction scores <0.5, suggesting in these cells my probabilities are split between two clusters. I am curious to see what the probabilities of the other 18 cells are for each of my cells, as currently I only have the prediction and mapping scores for the highest predicted cell type.

I wondered if it is possible to obtain the mapping and prediction score of the other 18 cell types for all of my cells also? Thanks for the help!

[andynkili]

Dear developers,

I also have the same issue where some cells have low score and it would help to compare the scores for all cell types from the reference, in order to have a broader picture of the annotation and understand those predicted low scores (some cell types from the reference being too similar?).

Best,
Andy

[andynkili]

@HarrietRB,

It is not obvious from the documentation but after a fair bit of googling, I got the scores from SeuratObject@assays$prediction.score.refMetadataUsedForAnnotation$data.
The [refMetadataUsedForAnnotation] is the name of metadata column from the reference that you have used for annotation.
Hope it helps.

Best,
Andy