FindTransferAnchors returns anchors object without anchors@weight.reduction, which cause error in TransferData().

[teryyoung]

hello,
my code is here,

# ...
query <- ScaleData(query, features = VariableFeatures(ref_data), verbose = FALSE)
query <- RunPCA(query, features=VariableFeatures(ref_data),npcs = 30)
# remove splitted scale.data layers
kept_scale_data <- GetAssayData(query, assay = "RNA", layer = "scale.data")
query <- DietSeurat(
    query, 
    assays = "RNA", 
    layers=c("counts","data"),
    dimreducs = names(query@reductions) 
)
query <- SetAssayData(query, assay = "RNA", layer = "scale.data", new.data = kept_scale_data)
ref_pca_features <- rownames(Loadings(ref_data, reduction = "pca"))

anchors <- FindTransferAnchors(
  reference = ref_data, 
  query = query, 
  dims = 1:12, 
  k.anchor = 10,
  reduction = "rpca", 
  # features = ref_pca_features, 
  reference.reduction = "integrated.rpca", 
  verbose = TRUE
)
predictions.clusters <- TransferData(anchorset = anchors, refdata = ref_data$ref_anno, dims=1:12, verbose = FALSE)
query <- AddMetaData(query, metadata = predictions.clusters)

my ref_data seems has no error:

> ref_data
An object of class Seurat 
14901 features across 18253 samples within 1 assay 
Active assay: RNA (14901 features, 1911 variable features)
 6 layers present: counts.2, data.2, scale.data.2, scale.data, counts, data
 4 dimensional reductions calculated: pca, integrated.rpca, umap.pca10, ref.umap_150n
> dim(Loadings(ref_data,reduction="integrated.rpca"))
[1] 1457   30

but FindTransferAnchors() return an anchors object with no weight.reduction:

> anchors@weight.reduction
A dimensional reduction object with key 
 Number of dimensions: 0 
 Number of cells: 0 
 Projected dimensional reduction calculated:  FALSE 
 Jackstraw run: FALSE 
 Computed using assay:  

how do I fix it?

[teryyoung]

appended:
I found the anchors returned by FindTransferAnchors() had some anchor pairs with NA:

> head(anchors_t@anchors)
     cell1 cell2     score
[1,]     1    NA 0.0000000
[2,]     2    NA 0.1428571
[3,]     3    NA 0.4285714
[4,]     4    NA 0.7142857
[5,]     5    NA 0.0000000
[6,]     6    NA 1.0000000
> length(anchors_t@anchors)
[1] 63282
> tail(anchors_t@anchors)
         cell1 cell2     score
[21089,] 18209 25347 0.0000000
[21090,] 18210 25348 1.0000000
[21091,] 18227 25365 1.0000000
[21092,] 18239 25378 0.5357143
[21093,] 18249 25388 0.5714286
[21094,] 18253 25393 0.5714286

is that normal?