Work with Visium HD

[alipirani88]

How do I bypass this error?

Error in h(simpleError(msg, call)) :
error in evaluating the argument 'x' in selecting a method for function 't': cannot allocate vector of size 722.0 Gb
Calls: SpaCET.deconvolution ... .V.rep.len -> .V.rep.int -> .handleSimpleError -> h
Execution halted

[beibeiru]

Hi, @alipirani88

Can you upload your SpaCET_obj to the following folder? I will fix it soon.

Best,
Beibei

[alipirani88]

Hi @beibeiru,

I an not able to access the link but I have uploaded the object to google drive:

[beibeiru]

Hi, @alipirani88

Thanks for uploading your data.

Your data is a Visium HD sample with 5,744,684 2µm-bins.
This is an extra-large matrix for deconvolution, causing memory issues.

It is better to aggregate your data to 16µm-bin so that you will get 5,744,684/64 = 89760 16µm-bins.
This has two advantages.
(1) a 16µm-bin has more expressed genes while a 2µm-bin has too many gene drop-out.
(2) a 2µm-bin is seldom able to cover an intact cell.

The solution for the current version of SpaCET package is that you can aggregate your data before your read your data as a SpaCET_object.
Alternatively, you can wait two days and I will upgrade the SpaCET package with an automatic aggregation function.

Best,
Beibei

[alipirani88]

Hi @beibeiru,

I get the following error with 16Mm bins:

Error in h(simpleError(msg, call)) : 
  error in evaluating the argument 'x' in selecting a method for function 't': unable to aggregate TsparseMatrix with 'i' and 'j' slots of length exceeding 2^31-1

Here is the link to object

https://drive.google.com/file/d/1h9j_JFMYjWt7exF77kAcfoK6SF3WmBTp/view?usp=sharing

[beibeiru]

Hi, @alipirani88

I fixed it and succeeded in running against your data on our HPC sever.

Please reinstall our package.

remove.packages(SpaCET)
devtools::install_github("data2intelligence/SpaCET")

There are many bins in your sample, which have a low count of expressed genes.
You can use min.genes to filter these low-quality bins, e.g., 50.

SpaCET_obj <- SpaCET.quality.control(SpaCET_obj, min.genes=50)
## [1] "Removing spots with less than 50 expressed genes."
## [1] "54865 spots are removed."
## [1] "91215 spots are kept."

SpaCET_obj <- SpaCET.deconvolution(SpaCET_obj, cancerType="BRCA", coreNo=8)

Best,
Beibei

[alipirani88]

Cool. It ran for me with 16Mm bins.

Does SpaCET work with Xenium data as well?

[beibeiru]

Xenium can capture hundreds of genes only.
I am not sure whether they are enough for SpaCET.
You can have a try and compare the results with H&E image.