subsetting object with multiple spatial fields of view fails with error

[pander21]

Hi, I ran into a bug that prevents subsetting the Seurat object with imaging-based spatial data that contains multiple different fields of view in images slot. Using subset on such object returns: Error in .nextMethod(x = x, i = i) : NAs not permitted in row index

see example code below:

# Two replicate sections with different cells, same genes
data1
An object of class Seurat 
302 features across 162939 samples within 1 assay 
Active assay: RNA (302 features, 0 variable features)
 1 spatial field of view present: fov_data1

data2
An object of class Seurat 
302 features across 154414 samples within 1 assay 
Active assay: RNA (302 features, 0 variable features)
 1 spatial field of view present: fov_data2

# Merge two sections
data_merged = merge(x = data1, y = data2)
An object of class Seurat 
302 features across 317353 samples within 1 assay 
Active assay: RNA (302 features, 0 variable features)
 2 spatial fields of view present: fov_data1 fov_data2

# Create classification column
data_merged$Groups = c(rep(c("Group1", "Group2"), ncol(data_merged)/2), "Group1")

# Try to subset the data
data_merged_subset = subset(data_merged, Groups == "Group1") # this fails with error

Error in .nextMethod(x = x, i = i) : NAs not permitted in row index

# When fields of view are removed, subset works again
data_merged@images$fov_data1 = NULL
data_merged@images$fov_data2 = NULL
data_merged_subset = subset(data_merged, Groups == "Group1")
data_merged_subset
An object of class Seurat 
302 features across 158677 samples within 1 assay 
Active assay: RNA (302 features, 0 variable features)

Same error occurs if Crop() is used to produce new FOV of one dataset and subsequently subset() is used on the whole object.

[ApolloWuu]

I've had the similar issue as well. I followed the spatial transcriptome vignette and created my object, but when I tried to subset the object based on the sum of counts for each cell, the same issue occurred: Error in .nextMethod(x = x, i = i) : NAs not permitted in row index.

The following is the script that I used:

vizgen.obj <- LoadVizgen(data.dir = getwd(), fov = "MERFISH_TEST1")

vizgen.obj

An object of class Seurat 135 features across 328851 samples within 1 assay Active assay: Vizgen (135 features, 0 variable features) 1 spatial field of view present: MERFISH_TEST1

vizgen.obj<- subset(vizgen.obj, subset = nCount_Vizgen > 0) Error in .nextMethod(x = x, i = i) : NAs not permitted in row index

[Daliya-K]

Same issue here, following!

[Daliya-K]

I think I have a workaround for my case. When I use LoadVizgen() I get an error in the subsetting of the loaded data: seur <- LoadVizgen(data.dir =sample_dir, fov=sample_name) Warning: 50956 cells missing polygon information

seur=subset(seur,subset = nCount_Vizgen >0) Error in .nextMethod(x = x, i = i) : NAs not permitted in row index

So I removed the cells for which there was no polygon information before adding the spatial info to the Seurat object: data <- ReadVizgen(data.dir = paste0(sample_dir), filter = "^Blank-", type = c("centroids", "segmentations"), z = 3L) data$centroids=data$centroids[data$centroids$cell %in%data$segmentations$cell,] data$transcripts=data$transcripts[,colnames(data$transcripts) %in%data$segmentations$cell]
segs <- CreateSegmentation(data$segmentations)
cents <- CreateCentroids(data$centroids) segmentations.data <- list(centroids = cents, segmentation = segs)
coords <- CreateFOV(coords = segmentations.data, type = c("segmentation", "centroids"), molecules = data$microns, assay = "Vizgen") seurObj <- CreateSeuratObject(counts = data$transcripts, assay = "Vizgen") coords <- subset(x = coords, cells = intersect(x = Cells(x = coords[["segmentation"]]), y = Cells(x = seurObj)))
seurObj[[sample_name]] <- coords

Now subsetting and normalizing of the seurat object works

[AustinHartman]

Hi, apologies for the slow reply - the bug in the FOV subset method should now be fixed on the feat/imaging branch.