SingleR basically takes normalized cell expression (expression values of all genes in each individual cell) as input

Step 1. Fetch reference annotation dataset from the Human Primary Cell Atlas (using celldex package)

ref <- celldex::HumanPrimaryCellAtlasData()

Step 2. Extract count data from the filtered Seurat object i.e. normalized cell expression

dta_counts <- GetAssayData(dta.seurat.filtered, slot = 'counts')

Step 3. Perform cell type annotation using SingleR by comparing test data to the reference dataset, assigning labels based on similarity

annotation <- SingleR(test = dta_counts, 
                      ref = ref, 
                      labels = ref$label.main)

Hope it helps!

• In order to run Azimuth:

AzimuthClusters <- RunAzimuth(seurat_obj, reference = "pbmcref")

• singleR

surveyReferences() # it lists all available references

ref <- fetchReference("monaco_immune", "2024-02-26")
seurat_dex <- as.SingleCellExperiment(seurat_obj)
predMainMonaco <- SingleR(test = seurat_dex, ref = ref, labels = ref$label.main, BPPARAM = BiocParallel::SerialParam())

Hope it helps

Perhaps you have tried this, but I used the Seurat function as.SingleCellExperiment() and this creates an object compatible with SingleR. DietSeurat() can be used to trim down the object to include only the features you want, just look up its documentation for more info.

Download your reference set from celldex for annotation

ref <- HumanPrimaryCellAtlasData(ensembl = TRUE)

Then create your SingleCellExperiment object

sce <- as.SingleCellExperiment(DietSeurat(seurat_obj))

Then you can perform your annotation

human.main <- SingleR(test = sce,assay.type.test = 1,ref = ref,labels = ref$label.main)

This resource might be helpful, it goes into using SingleR in more depth.

Perhaps you can try this tool: https://omnibusx.com/prediction