Analyse stereoseq from TMA with several core

[synatkeamsk]

Dear Giotto team,

Hope you are well. I'm currently analyzing stereoseq spatial transcriptomic data from one TMA with several cores. I am wondering whether Giotto has pipeline for TMA from stereoseq?

Also, I have installed Giotto suit latest version 4.2.2. However, I could not get the function createGiottoStereoSeqObject. Could you kindly help me how to get this function installed. Also, for full analysis, may I run via Rstudio on Window or you think I should entirely run via terminal? I have a powerful computer too on window. but I heard that Giotto is more robust with terminal.
I hope you could help.

Thanks,

Synat,

[josschavezf]

Hi @synatkeamsk, the createGiottoStereoSeqObject() function is currently in the development branch of Giotto (we will move it to the main one during this week), You can install the dev version using the following commands:

remotes::install_github("drieslab/GiottoUtils@dev")
remotes::install_github("drieslab/GiottoClass@dev")
remotes::install_github("drieslab/GiottoVisuals@dev")
remotes::install_github("drieslab/Giotto@suite_dev")

We don't have a specific tutorial for TMA, but the function should in theory work for any stereoseq dataset, please let us know if you find any issue while running it with your sample.

Depending on the RAM of your computer, the analysis could run using RStudio locally when reading the data at the cell level or the big bin sizes (100 or 50), but it will need lot of memory when using small bin sizes (5, 2, or 1) or reading the subcellular data, which may require a powerful computer or even a cluster.

[josschavezf]

@synatkeamsk following up in this topic. It was just brought to my attention that we do have a tutorial for TMA analysis https://giottosuite.com/articles/tma.html, it's not specific for stereoseq but in theory it should work too for this technology.

[synatkeamsk]

Dear Joselyn,

Thanks for your kind support and guidance. After carefully following your guidance, I have successfully installed the development branch of Giotto and got createGiottoStereoSeqObject working.

Thanks for sharing TMA tutorial. I hope it will run smoothly for my TMA. I will reach out again if I got stuck. Hope you do not mind.

Kind Regards,

synat

[synatkeamsk]

Hi Joselyn,

I tested it and got a bit of error. I have the data folder and structure exactly the same as the structure provided by Giotto. It says Assertion on 'raster_object' failed: No file provided. You can help review and responded whenever you have time. Appreciate! I attached the content in my image folder. It looks like I did not have H&E image there at all. Not sure why because these are the image files I got from Sequencing facility who did the sequencing and alignment. What do you think about it? Do I need H&E or whether the all the images in the attached image will work .

Also, I just had a chat with Bioinformatician who did the alignment and got the response like this "The current version doesn't include an H&E image because it was run using the single cell segmentation version using a dapi like stain". Therefore, I cannot provide an H&E image for this". Is there a way to get around with the new current version that does not include H&E using createGiottoStereoSeqObject. Hope you could help again.

library(Giotto)

# Set the results directory to save plots
results_folder <- "Z:/Single_cell_omic/Stomic_stereoseq/Stereoseq-thyroid/Results"

# Optional: Specify the path to a Python executable within a conda or miniconda 
# environment. If set to NULL (default), the Python executable within the previously installed Giotto environment will be used.
python_path <- NULL # alternatively, "/local/python/path/python" if desired.

# Create Giotto Instructions
instructions <- createGiottoInstructions(save_dir = results_folder,
                                         save_plot = TRUE,
                                         show_plot = FALSE,
                                         return_plot = FALSE,
                                         python_path = python_path)


data_path <- "Z:/Single_cell_omic/Stomic_stereoseq/Stereoseq-thyroid/MM_NA_IMG_HD_Y01679G5"

## read the data
Giotto.obj.bin100 <- createGiottoStereoSeqObject(stereoseq_dir = data_path,
                               type = "squarebin",
                               bin_size = "bin100",
                               gene_column = "geneName",
                               negative_y = TRUE,
                               instructions = instructions)

Reading expression file...
Finished reading in tissue.gef
Creating spatial_locations...
12709 bins in total
Finished spatial_locations
Creating expression matrix...
Processed 18275982 groups out of 18275982. 100% done. Time elapsed: 110s. ETA: 0s.
finished expression matrix
Creating giotto object...
Setting expression [cell][rna] raw
--- finished expression data ---

Setting spatial locations [cell] raw
--- finished spatial location data ---

--- finished cell metadata ---

--- finished feature metadata ---

No spatial network results are provided
No spatial enrichment results are provided
No dimension reduction results are provided
No nearest neighbor network results are provided
Attaching HE image...
Error in createGiottoLargeImage(he_image_path, name = "image", negative_y = negative_y) :
Assertion on 'raster_object' failed: No file provided.

Thanks,

synat

[josschavezf]

@synatkeamsk thanks for reporting this issue, I just make a quick fix. Please install again Giotto from the dev branch, no need to re-install the other modules, only run this command:

remotes::install_github("drieslab/Giotto@suite_dev")

[synatkeamsk]

Dear Joselyn,

Thanks for the update. I will reinstall and run it again.

I will also try TMA tutorial. Hopefully you can help guide if I encounter any issue! I really appreciated your support!

Kind Regards,

Synat

[synatkeamsk]

Hi Jocelyn,

I have tested the new function that you had fixed. However, I realized that with the new version that based on "runing using the single cell segmentation version using a dapi like stain" , the BinType is "CellBin", not a squarebin. I am wondering whether createGiottoStereoSeqObject still work for type= "cellbin". or if it worked, how to specify the parameter in createGiottoStereoSeqObject.

g <- createGiottoStereoSeqObject(stereoseq_dir = data_path,
                               type = "squarebin",
                               bin_size = "bin100",
                               gene_column = "geneName",
                               negative_y = TRUE,
                               instructions = instructions)

I tired to run the following, but it did not work and got error

g <- createGiottoStereoSeqObject(stereoseq_dir = data_path,
                               type = "cellbin",
                               gene_column = "geneName",
                               negative_y = TRUE,
                               instructions = instructions)

Reading expression file...
Finished reading in adjusted.cellbin.gef
Creating spatial_locations...
308140 cells in total
Finished spatial_locations
Creating expression matrix...
Processed 24435357 groups out of 24435357. 100% done. Time elapsed: 107s. ETA: 0s.
Error in (function (..., sorted = TRUE, unique = FALSE) :
Cross product of elements provided to CJ() would result in 13876468620 rows which exceeds .Machine$integer.max == 2147483647

In the analysis folder under outs folder, I still see bin20 and bin50 file (image attach). I hope you could help with this one too because I still stuck.

Kind Regards

[josschavezf]

Hi @synatkeamsk this last error is actually not related to the organization of the files. I found this error recently and it's related to the size of the data table when creating the gene x cell/bin expression matrix, which seems to exceed the computational limits in R, it keeps happening even when using a big computer because it's not related to the RAM capabilities, but to limitations in the language. I'm still working in a solution and hope to solve it during the next week. Apologies for the delay.

[synatkeamsk]

Dear Jocelyn,

Thanks for your kind support. Kindly help me address if the Bin Type is not squarebin and how I could change the parameter in the function below if the binType is cellbin. No worries, you can respond whenever you have time.

g <- createGiottoStereoSeqObject(stereoseq_dir = data_path,
                               type = "squarebin",
                               bin_size = "bin100",
                               gene_column = "geneName",
                               negative_y = TRUE,
                               instructions = instructions)

Kind Regards,

synat,

[synatkeamsk]

HI Jocelyn,

Hope you are well. I am wondering whether the issue has been fixed.

Kind Regards,

Synat