anndataR ignores outcomes of SCT

[LouiseDck]

After running SCT on a Seurat dataset you end up with 2 issues:

• scale.data has less features and can't easily be transferred
• there are these extra "SCTModel"s added to the Seurat object, which don't seem to fit cleanly in any of the anndata slots.

Do we want to do anything with this? Do we want to provide a vignette on how to deal with this?

reprex:

library(Seurat)
#> Loading required package: SeuratObject
#> Loading required package: sp
#> 
#> Attaching package: 'SeuratObject'
#> The following objects are masked from 'package:base':
#> 
#>     intersect, t
library(anndataR)
library(sctransform)
library(SeuratData)
#> ── Installed datasets ──────────────────────────────── SeuratData v0.2.2.9002 ──
#> ✔ pbmc3k 3.1.4
#> ────────────────────────────────────── Key ─────────────────────────────────────
#> ✔ Dataset loaded successfully
#> ❯ Dataset built with a newer version of Seurat than installed
#> ❓ Unknown version of Seurat installed

InstallData("pbmc3k")
#> Warning: The following packages are already installed and will not be
#> reinstalled: pbmc3k
pbmc <- UpdateSeuratObject(pbmc3k)
#> Validating object structure
#> Updating object slots
#> Ensuring keys are in the proper structure
#> Warning: Assay RNA changing from Assay to Assay
#> Ensuring keys are in the proper structure
#> Ensuring feature names don't have underscores or pipes
#> Updating slots in RNA
#> Validating object structure for Assay 'RNA'
#> Object representation is consistent with the most current Seurat version

# store mitochondrial percentage in object meta data
pbmc <- PercentageFeatureSet(pbmc, pattern = "^MT-", col.name = "percent.mt")

# run sctransform
pbmc <- SCTransform(pbmc, vars.to.regress = "percent.mt", verbose = FALSE)
#> Warning: The `slot` argument of `GetAssayData()` is deprecated as of SeuratObject 5.0.0.
#> ℹ Please use the `layer` argument instead.
#> ℹ The deprecated feature was likely used in the Seurat package.
#>   Please report the issue at <https://github.com/satijalab/seurat/issues>.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.
#> `vst.flavor` is set to 'v2' but could not find glmGamPoi installed.
#> Please install the glmGamPoi package for much faster estimation.
#> --------------------------------------------
#> install.packages('BiocManager')
#> BiocManager::install('glmGamPoi')
#> --------------------------------------------
#> Falling back to native (slower) implementation.
#> Warning in theta.ml(Y, mu, sum(w), w, limit = control$maxit, trace =
#> control$trace > : iteration limit reached
#> ...
#> Warning: The `slot` argument of `SetAssayData()` is deprecated as of SeuratObject 5.0.0.
#> ℹ Please use the `layer` argument instead.
#> ℹ The deprecated feature was likely used in the Seurat package.
#>   Please report the issue at <https://github.com/satijalab/seurat/issues>.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.

adata <- from_Seurat(pbmc, layers = c("counts", "data"))

dim(LayerData(pbmc, "counts"))
#> [1] 12572  2700
dim(LayerData(pbmc, "data"))
#> [1] 12572  2700
dim(LayerData(pbmc, "scale.data"))
#> [1] 3000 2700

pbmc[["SCT"]]@SCTModel.list
#> $model1
#> An sctransform model.
#>   Model formula:  y ~ log_umi 
#>   Parameters stored for 12572 features, 2700 cells.

adata
#> AnnData object with n_obs × n_vars = 2700 × 12572
#>     obs: 'orig.ident', 'nCount_RNA', 'nFeature_RNA', 'seurat_annotations', 'percent.mt', 'nCount_SCT', 'nFeature_SCT'
#>     layers: 'counts', 'data'

Created on 2025-05-20 with reprex v2.1.1

[lazappi]

I don't think the SCTModels matter too much because you won't really be able to do much with them in the anndata anyway.

Maybe if there are layers with different sizes we should only convert the smallest number of features with a warning? Or have an argument that controls something like that?