Behaviour of MapQuery for multimodal query using multimodal reference

[jasonyklam]

I have generated a multiome reference dataset from two healthy control samples (similar way to #5346, in brief, integrating RNA and ATAC assay independently, then calculate WNN by FindMultiModalNeighbors). I integrated the RNA assay using Seurat v4 approach so it returned an 'integrated' assay which contained a single SCT model that I could use for reference mapping directly. My problem is now I wanted to map a patient sample using the above reference dataset. The sample is also a multiome dataset (RNA + ATAC). I've written the code below to do it as suggested in #6537:

DefaultAssay(integrated_small) <- "integrated"       
integrated_small <- RunSPCA(integrated_small, graph = 'wsnn')

anchors <- FindTransferAnchors(reference = integrated_small, 
                               query = sample_5, 
                               normalization.method = "SCT", 
                               reference.reduction = "spca",
                               dims = 1:50)

sample_5 <- MapQuery(anchorset = anchors, 
                                 query = sample_5, 
                                 reference = integrated_small,
                                 refdata = list(annotation = "annotation",
                                 clusters = "seurat_clusters"), 
                                 reference.reduction = "spca", 
                                 reduction.model = "wnn.umap")                                

This gave me the predicted annotation and cluster labels, but I'm not sure what information is used from the reference/query object to generate the results. According to the multimodal mapping tutorial -
(https://satijalab.org/seurat/articles/multimodal_reference_mapping)
it stated that MapQuery is a wrapper around three functions: TransferData(), IntegrateEmbeddings(), and ProjectUMAP(). The TransferData() function seems to impute values of the missing assay (ADT in the example). Since my query object is already multimodal (RNA + ATAC), would the MapQuery function still try to impute the ATAC assay given that it is already present?

Also as for the output, were the predictions that it returned based on the RNA assay of the query only? Or the original/imputed ATAC assay was also taken into account? If so then would the imputed values replace the original ones?

I saw #8925 having a similar issue with CITE-seq dataset but the answer is unclear.

I'm new to bioinformatics so it would be great if someone could explain things in simple terms...!

Thank you!