Using ATAC-bridge in RunAzimuth()

[dy-lin]

Is there a way to obtain celltype labels for an ATAC human bonemarrow dataset using RunAzimuth()? Is there an ATAC-bridge reference in SeuratData?

Or does a custom reference have to be built using the h5ad files (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE194122)?

[Gesmira]

Hi, We haven't uploaded the ATAC references to SeuratData yet, but you can use the Azimuth web app or find the bone marrow reference here to download and run locally in R with RunAzimuth().

[dy-lin]

Is the idx.annoy file the same as the one for RNA? https://zenodo.org/record/5521512#.ZGPHyXbMKJE

[dy-lin]

Using that idx.annoy, I get:

> list.files()
[1] "idx.annoy" "ref.Rds"  
> atac.ref <- RunAzimuth(
+     atac,
+     reference = "."
+ )
Error: Cannot find 'refdr.annoy.neighbors' in this Seurat object

[Gesmira]

Hi, For mapping scATAC-seq data, please refer to the following vignette.

# Specify query.modality and fragment.path for ATAC data 
atac <- RunAzimuth(query = "~/10k_PBMC_ATAC_nextgem_Chromium_X_filtered_peak_bc_matrix.h5", query.modality = "ATAC",
    reference = "~/reference", fragment.path = "~/10k_PBMC_ATAC_nextgem_Chromium_X_fragments.tsv.gz")

[dy-lin]

My scATAC-seq object is an integrated object (integrated using IntegrateEmbeddings()). Is it advisable to just cat the fragment files (with updated barcode names to correspond to integrated samples) for the samples that have been integrated?

[dy-lin]

After some manipulation of each individual atac fragments file to match the barcode names to the integrated object (i.e. ${sample_name}_${barcode}), I'm unable to run tabix on the fragments file.

$ tabix -p bed all.atac_fragments.body.sorted.tsv.gz
tbx_index_build failed: all.atac_fragments.body.unix_sorted.tsv.gz