How to introduce normalized and scaled seurat data into monocle 3?

Question

[Sun]

Hello every,

I am trying to transform a seurat oject to a monocle 3 object. I use the following codes:

seur_obj <- NormalizeData(seur_obj)
seur_obj <- FindVariableFeatures(seur_obj)
seur_obj <- ScaleData(seur_obj)
seur_obj <- RunPCA(seur_obj)
seur_obj <- RunUMAP(seur_obj, dims = 1:20)

data <- GetAssayData(seur_obj, assay = 'RNA', slot = 'counts')
cell_metadata <- [e-mail]
gene_annotation <- data.frame(gene_short_name = rownames(data))
rownames(gene_annotation) <- rownames(data)
mono_obj <- new_cell_data_set(data, cell_metadata = cell_metadata, 
                        gene_metadata = gene_annotation)

It seems that only the "counts" slot in the seur_obj can be introduced into the monocle object, and is stored in the "@assays@data$counts" slot. If I use "scale.data" in the

"GetAssayData" function like:
data <- GetAssayData(seur_obj, assay = 'RNA', slot = 'scale.data')
cell_metadata <- [e-mail]
gene_annotation <- data.frame(gene_short_name = rownames(data))
rownames(gene_annotation) <- rownames(data)
mono_obj <- new_cell_data_set(data, cell_metadata = cell_metadata, 
                        gene_metadata = gene_annotation)

I will get error:

Error in log(cell_total) : (converted from warning) NaNs produced

So how to introduce normalized and scaled seurat data into mono_obj? Or is there functions to do normalization and scaling in monocle?

In Monocle 3, the function "preprocess_cds" normalize and scale the data by default, by I don't know where the nornalized and scaled data is stored. I can only find the counts data in mono_obj@assays@data$counts.

Beside, I want to do some analysis on some clusters based on the normalized scaled data in the mono_obj. Because I can only find the count data, I expect to convert the mono_obj to a Seurat object. In monocle 2, the function "exportCDS" can do this, but this function has disappeared. Is there any other function that does this in monocle 3?

Thank you in advance!

[fracarb8]

Have a look at as.cell_data_set from SeuratWrappers

Answer 1

[Reza]

I have a same question it seems that when we use the preprocess_cds() function, it does not save the normalized data directly in a separate slot for general use. Instead, normalization is typically applied on-the-fly for dimensionality reduction and trajectory analysis purposes. I want to use graph_test() but not sure if should manually normalize the data first (in Seurat ) before calling the graph_test() function.

[Bastien Hervé]

For OP's question, my guess is that they got NA's because ScaleData is only called on the VariableFeatures found the step before.

Try : seur_obj <- ScaleData(seur_obj, features=rownames(seur_obj))

For your question, you should be able to do both, in Monocle you can use normalize_expr_data to normalize your matrix.

Also helpful : https://github.com/cole-trapnell-lab/monocle-release/issues/162