Subset visium-hd dataset without removing other binning data

[dahun73]

Hello :)

I'm trying to analyze visium-hd data using Seurat. In tutorial, there is no filtering step for removing low quality bins. Because there are several low quality bins in my sample, I should filter those bins now.
When I filtered the bins following parameters(nCount_RNA, nFeature_RNA and percent.mt) using 'subset' function, it removed other binning data.

> hd_obj2_sel1 <- subset(hd_obj2,subset =  percent.mt_Spatial.008um < 10 & nCount_Spatial.008um > 100 & nFeature_Spatial.008um > 50)
Spatial.002um assay doesn't leave any cells, so it is removed
Spatial.016um assay doesn't leave any cells, so it is removed

Of course, I can try to analyze again using other bins. However, I want to know how to leave other assays. If not, are there any suggestions for filtering low quality bins?

Thank you in advance :)

[jdrnevich]

We had the same issue. I don't know what the solution is, but the problem is that the metadata in the Seurat object contains one row for each binning level/bin, but different columns for nCount_Spatial.008um and nCount_Spatial.016um, and there are NA values for the opposite nCount:

> Con_324_left <- Load10X_Spatial("results/spaceranger/A324-S5-left/outs/", 
+                                 slice = "Con_324_left", bin.size = c(8, 16))
> Con_324_left
An object of class Seurat 
38118 features across 184501 samples within 2 assays 
Active assay: Spatial.008um (19059 features, 0 variable features)
 1 layer present: counts
 1 other assay present: Spatial.016um
 2 spatial fields of view present: Con_324_left.008um Con_324_left.016um
> head(Con_324_left)
                      orig.ident nCount_Spatial.008um nFeature_Spatial.008um nCount_Spatial.016um
s_008um_00773_00596-1          s                  407                    192                   NA
s_008um_00269_00526-1          s                  652                    258                   NA
s_008um_00464_00652-1          s                   70                     67                   NA
s_008um_00547_00611-1          s                  373                    341                   NA
s_008um_00693_00628-1          s                  327                    280                   NA
s_008um_00466_00453-1          s                   29                     26                   NA
s_008um_00433_00599-1          s                  383                    346                   NA
s_008um_00660_00606-1          s                  238                    212                   NA
s_008um_00169_00602-1          s                  491                    405                   NA
s_008um_00622_00588-1          s                  439                    375                   NA
                      nFeature_Spatial.016um
s_008um_00773_00596-1                     NA
s_008um_00269_00526-1                     NA
s_008um_00464_00652-1                     NA
s_008um_00547_00611-1                     NA
s_008um_00693_00628-1                     NA
s_008um_00466_00453-1                     NA
s_008um_00433_00599-1                     NA
s_008um_00660_00606-1                     NA
s_008um_00169_00602-1                     NA
s_008um_00622_00588-1                     NA
> tail(Con_324_left)
                      orig.ident nCount_Spatial.008um nFeature_Spatial.008um nCount_Spatial.016um
s_016um_00282_00226-1          s                   NA                     NA                    7
s_016um_00325_00338-1          s                   NA                     NA                 2078
s_016um_00370_00214-1          s                   NA                     NA                    4
s_016um_00086_00272-1          s                   NA                     NA                  912
s_016um_00320_00280-1          s                   NA                     NA                  273
s_016um_00176_00310-1          s                   NA                     NA                  886
s_016um_00167_00246-1          s                   NA                     NA                 1212
s_016um_00355_00284-1          s                   NA                     NA                 1277
s_016um_00371_00313-1          s                   NA                     NA                 1581
s_016um_00323_00266-1          s                   NA                     NA                  123
                      nFeature_Spatial.016um
s_016um_00282_00226-1                      7
s_016um_00325_00338-1                   1557
s_016um_00370_00214-1                      4
s_016um_00086_00272-1                    469
s_016um_00320_00280-1                    193
s_016um_00176_00310-1                    726
s_016um_00167_00246-1                    663
s_016um_00355_00284-1                    951
s_016um_00371_00313-1                   1126
s_016um_00323_00266-1                    113
> nrow(Con_324_left[[]])
[1] 184501
> ncol(Con_324_left[["Spatial.016um"]]$counts)
[1] 37178
> ncol(Con_324_left[["Spatial.008um"]]$counts)
[1] 147323
> 147323 + 37178
[1] 184501

Maybe a hack would be to replace the NAs with the minimum value to be kept at the other level? So on your data something like:

hd_obj2_sel1$nCount_Spatial.008um[is.na(hd_obj2_sel1$nCount_Spatial.008um)] <- 101

[dahun73]

Hi @jdrnevich,

As you have seen, the barcodes are all different depending on the bin size, so in my opinion, it would be best to create different objects for each bin size in order to filter and analyze them.

[jdrnevich]

We had just come to the same conclusion - different objects for different bin sizes! The bin switching in one object was an interesting idea but won't work if all bins are in one meta data object