FindTransferAnchors error: "No features to use in finding transfer anchors" when using CELLxGENE .h5ad reference with Seurat

[snijesh]

I downloaded the dataset from the study “A single-cell and spatially resolved atlas of human breast cancers” from CELLxGENE:

https://cellxgene.cziscience.com/collections/dea97145-f712-431c-a223-6b5f565f362a

Download command:

wget -c https://datasets.cellxgene.cziscience.com/0e3974a1-9488-4987-9767-21d4e389fac7.h5ad

I converted the .h5ad file into a Seurat object and attempted to use it as a reference for label transfer.

Code used:

library(Seurat)
library(sceasy)
library(schard)

# load h5ad as Seurat
reference = schard::h5ad2seurat("0e3974a1-9488-4987-9767-21d4e389fac7.h5ad")

gene_symbols <- reference@assays$RNA@meta.features$feature_name

# replace empty values
gene_symbols[gene_symbols == "" | is.na(gene_symbols)] <- rownames(reference)
# update assay matrix rownames
rownames(reference@assays$RNA@counts) <- make.unique(gene_symbols)
rownames(reference@assays$RNA@data) <- make.unique(gene_symbols)

# update Seurat object rownames
rownames(reference) <- make.unique(gene_symbols)

tma1 = readRDS("tma1_umap.rds")

anchors <- FindTransferAnchors(reference = reference, query = tma1, dims = 1:20)

However, the following error occurs:

Error: No features to use in finding transfer anchors. 
To troubleshoot, try explicitly providing features to the features parameter and ensure that they are present in both reference and query assays.

Reference object:

28468 features across 100064 cells

Query object (tma1):

18065 features across 64665 cells

Has anyone successfully used this CELLxGENE dataset as a Seurat reference for label transfer?
What could be causing the FindTransferAnchors error in this case?

[longmanz]

Hi,
Have you checked the feature names in your tma1_umap.rds? You can run a quick intersect between the feature names of your reference and your own object to see if they have the same naming scheme.

In addition, you should run standard preprocessing like NormalizeData, FindVariableFeatures, ScaleData, and RunPCA for your reference as well as your own object (for example, see: https://satijalab.org/seurat/reference/findtransferanchors).