changes of coordinates using convert.Seurat() function

[SoniaHermoso]

Dear developers,

I have just updated the last version of SpaCE and when I have run new samples I have found that the coordinates are not correct.

This is the code:

SpaCET_obj <- convert.Seurat(SpaCET_obj , platform = "Visium")
SpaCET_obj <- SpaCET.deconvolution(SpaCET_obj , cancerType="PRAD", coreNo=8)
SpaCET.visualize.spatialFeature(
SpaCET_obj ,
spatialType = "CellFraction",
spatialFeatures="All",
sameScaleForFraction = F,
pointSize = 0.5,
nrow=6)

I did not have this issue with the old version.

Do I need to use the argument "visiumPath" in convert.Seurat() function to fix it?

Thank you!
Sonia H.

[beibeiru]

Hi, @SoniaHermoso ,

Thank you for reporting this bug!

Would you please share your Seurat object before running convert.Seurat? I will fix it soon.

Best,
Beibei

[SoniaHermoso]

Hi Beibei,

I've been comparing two Seurat objects with their respective SpaCET objects in detail to see if I could find any differences. One of the objects has correct coordinates (we'll call it "obj1") and the other has incorrect coordinates (we'll call it "obj2").

head(obj1_Seurat@images[["slice1"]]@boundaries[["centroids"]]@coords)
x y
[1,] 2991.149 1043.147
[2,] 5670.170 3844.672
[3,] 2153.865 1699.044
[4,] 4116.144 3657.125
[5,] 6481.858 5691.315
[6,] 4012.931 3462.268
obj1_Seurat@images[["slice1"]]@scale.factors[["lowres"]]
[1] 0.05923586
head(obj1_SpaCET@input[["spotCoordinates"]])
pxl_row pxl_col barcode
61.7917334836277x177.183274616061 61.79173 177.1833 AACACGTGCATCGCAC-1
227.7424590516x335.877376084662 227.74246 335.8774 AACACTTGGCAAGGAA-1
100.644319662244x127.586022269527 100.64432 127.5860 AACAGGAAGAGCATAG-1
216.632909208663x243.82331806528 216.63291 243.8233 AACAGGATTCATAGTT-1
337.129906953792x383.958427202698 337.12991 383.9584 AACAGGTTATTGCACC-1
205.090406086915x237.709400726979 205.09041 237.7094 AACAGGTTCACCGAAG-1

This means that “pixel_row” corresponds to the “y” coordinate, and “pixel_col” to the “x” coordinate in obj1 (correct).

In obj2(not correct), I observe the opposite: “pixel_row” corresponds to the “x” coordinate and vice versa.

I went to your public R code and saw this when you define spotCoordinates:
…
spotCoordinates <- data.frame(
pixel_row=slice@boundaries$centroids@coords[,1] * slice@scale.factors$lowres,
pixel_col=slice@boundaries$centroids@coords[,2] * slice@scale.factors$lowres
)
…

If I understand correctly, you should change to:
…
spotCoordinates <- data.frame(
pixel_col=slice@boundaries$centroids@coords[,1] * slice@scale.factors$lowres,
pixel_row=slice@boundaries$centroids@coords[,2] * slice@scale.factors$lowres
)
…

Do you think the bug could be here?

Sonia H.

[beibeiru]

Hi, Sonia,

The function SpaCET.visualize.spatialFeature already includes a flip option, so my code should be correct.
The example below produces the expected figure:

library(SpaCET)
visiumPath <- file.path(system.file(package = "SpaCET"), "extdata/Visium_BC")
Seurat_obj <- Seurat::Load10X_Spatial(data.dir = visiumPath)
SpaCET_obj <- convert.Seurat(Seurat_obj, platform = "Visium")
SpaCET_obj <- SpaCET.quality.control(SpaCET_obj)
SpaCET.visualize.spatialFeature(SpaCET_obj, spatialType = "QualityControl", spatialFeatures = c("UMI","Gene"))

If possible, you’re welcome to share your data with me (beibei.ru@nih.gov). I will keep it confidential and remove it immediately after resolving the issue.

Best,
Beibei

[SoniaHermoso]

Hi Bebibei,

I've checked the coordinates again. To do this, I repeated the convert.Seurat() function on the correct object (previously called "obj1") and the plot shows incorrect coordinates. However, in the previous version of SpaCET, they worked correctly. I think something changed when I updated the SpaCET version. The Seurat version was also recently updated.

New SpaCET version (incorrect coordinates)

Previous SpaCET version (correct coordinates)

Sonia H.

[ahwanpandey]

I m having the same issue here with my data after doing "convert.Seurat"

[ahwanpandey]

Doing this worked for me, basically swithing row and column coords:

  SpaCET_obj <- SpaCET::convert.Seurat(seurat_object, platform = "Visium")
  
  a = SpaCET_obj@input$spotCoordinates$pixel_row
  b = SpaCET_obj@input$spotCoordinates$pixel_col
  SpaCET_obj@input$spotCoordinates$pixel_row <- b
  SpaCET_obj@input$spotCoordinates$pixel_col <- a
  
  SpaCET_obj <- SpaCET.quality.control(SpaCET_obj, min.genes=100)
  
  # plot the QC metrics
  SpaCET.visualize.spatialFeature(
    SpaCET_obj, 
    spatialType = "QualityControl", 
    spatialFeatures=c("UMI","Gene"),
    imageBg = TRUE,
    imageSize = "CompleteImage"
  )