Requested feature: handle inputs from Seurat #262

TomKellyGenetics · 2019-11-26T01:24:00Z

Proposal to implement to extract a cell_data_set from a Seurat object. Can submit a PR if needed.

diegoalexespi · 2019-12-11T20:34:42Z

hi @TomKellyGenetics - I've been able to extract a monocle3 CDS from a seurat v3 object - would be happy to share the function here or as part of a pull request, I'm not sure which is best. Granted, it's for moving a SCT & harmony-integrated dataset from Seurat to monocle3 so it may not work with all input Seurat objects

TomKellyGenetics · 2019-12-23T04:06:25Z

@d93espinoza That would be great. I've managed to extract the relevant fields of a Seurat object but haven't made it into a function (thinking about the best way to do this). If you've already done it, that would be great, I hope to see it as a PR.

summerrfair · 2020-01-15T02:37:22Z

@d93espinoza Hello! I am hoping to perform pseudotime analysis on an integrated Seurat (v3) object in Monocle. It sounds like you have performed similar analyses - I would greatly appreciate your help and any code you may be willing to share. Thank you so much!

TomKellyGenetics · 2020-01-15T05:11:29Z

@summerrfair Based on the documentation here, this is a working example on a Suerat object. Note that monocle takes the raw counts (merged or integrated objects are okay) and requires the gene names and barcodes to match the metadata.

library("org.Hs.eg.db")
gene_symbol <- as.list(org.Hs.egSYMBOL)
library("Suerat")
library("monocle3")
corrected_data <- GetAssayData(sample.integrated, assay = "integrated", slot = "data")
raw_count_data <- GetAssayData(sample.integrated, assay = "RNA", slot = "counts")
class(raw_count_data)

cells_info <- sample.integrated@meta.data

gene_name <- gene_symbol[rownames(raw_count_data)]  
gene_name <- sapply(gene_name, function(x) x[[1]][1])

#preparing cds
gene_name <- ifelse(is.na(gene_name), names(gene_name), gene_name)
gene_short_name <- gene_name
gene_id <- rownames(raw_count_data)


genes_info <- cbind(gene_id, gene_short_name)
genes_info <- as.data.frame(genes_info)
rownames(genes_info) <- rownames(raw_count_data)

cds <- new_cell_data_set(expression_data = raw_count_data,
                         cell_metadata = cells_info,
                         gene_metadata = genes_info)
cds

You can also ensure that the clusters and dimension reductions match those computed in Suerat like this:

#replace monocle clusters with seurat
cds@clusters$UMAP$partitions <- sample.integrated@meta.data$seurat_clusters
names(cds@clusters$UMAP$partitions) <- rownames(sample.integrated@meta.data)
cds@clusters$UMAP$clusters <- sample.integrated@meta.data$seurat_clusters
names(cds@clusters$UMAP$clusters) <- rownames(sample.integrated@meta.data)

#replace monocle dimensions with seurat
cds@reducedDims$TSNE <-  suerat_object@reductions$tsne@cell.embeddings
cds@reducedDims$UMAP <-  suerat_object@reductions$umap@cell.embeddings

It wouldn't be too difficult to wrap either of these into a function but not sure on the best way to do this.

summerrfair · 2020-01-16T21:59:28Z

@TomKellyGenetics

Many thanks for sharing this. Very helpful! I will try to implement and will let you know how it goes.

diegoalexespi · 2020-02-24T20:02:51Z

Here's my try (so far) at the function, sorry for the delay and thanks @TomKellyGenetics for sharing your code!

ToMonocle3 <- function(seurat_object,
                       scale_all = FALSE,
                       assay = "SCT",
                       reduction_for_projection = "PCA",
                       UMAP_cluster_slot = NULL){
  
  if(scale_all){
    message("Getting residuals for all Seurat genes in chosen assay slot and placing in scale.data")
    seurat_genes <- rownames(seurat_object[[assay]])
    remaining_genes <- setdiff(seurat_genes, rownames(seurat_object[[assay]]@scale.data))
    if(assay == "SCT"){
      seurat_object <- Seurat::GetResidual(seurat_object, features = remaining_genes, assay = assay, umi.assay = "RNA")
    } else {
      seurat_object <- Seurat::ScaleData(seurat_object, features = rownames(seurat_object[[assay]]))
    }
  }
  
  #We prep the seurat object by creating gene loadings for ALL genes in the Seurat scale.data slot. This is done to allow downstream monocle3 functions on gene_modules to work appropriately.
  message("Projecting gene loadings for all Seurat genes in scale.data slot")
  seurat_object <- Seurat::ProjectDim(seurat_object, reduction = reduction_for_projection, assay = assay)
  
  ##################
  
  message("Initializing CDS object")
  
  #Extract Seurat's log-transformed values
  expression_matrix <- Seurat::GetAssayData(seurat_object, assay = assay, slot = "counts")
  #Extract Seurat meta_data
  meta_data <- seurat_object@meta.data
  #Extract gene names from Seurat object SCT slot to make CDS
  seurat_genes <- data.frame(gene_short_name = rownames(seurat_object[[assay]]),
                             row.names = rownames(seurat_object[[assay]]))
  new_cds <- monocle3::new_cell_data_set(expression_data = expression_matrix, cell_metadata = meta_data, gene_metadata = seurat_genes)
  
  ##################
  
  message("Making an SCE object from the Seurat object to facilitate transfer of information from SCE to CDS")
  sce <- as.SingleCellExperiment(seurat_object, assay = assay)
  message("Loading in all Seurat reductions (PCA, HARMONY, UMAP, etc.) into CDS")
  SingleCellExperiment::reducedDims(new_cds) <- SingleCellExperiment::reducedDims(sce)
  message("Loading in specified Seurat assay into CDS")
  SummarizedExperiment::assays(new_cds) <- SummarizedExperiment::assays(sce)
  message("Loading in Seurat gene names into CDS")
  SummarizedExperiment::rowData(new_cds) <- SummarizedExperiment::rowData(sce)
  SummarizedExperiment::rowData(new_cds)$gene_short_name <-  row.names(new_cds)
  message("Loading in Seurat gene loadings into CDS")
  new_cds@preprocess_aux$gene_loadings <- seurat_object@reductions[[reduction_for_projection]]@feature.loadings.projected
  
  ##################
  
  message("Get user specified selected clusters (or active idents) from Seurat and load into CDS")
  if(is.null(UMAP_cluster_slot)){
    list_cluster <- Idents(seurat_object)
  } else {
    Idents(seurat_object) <- UMAP_cluster_slot
    list_cluster <- Idents(seurat_object)
  }
  new_cds@clusters[["UMAP"]]$clusters <- list_cluster
  #The next two commands are run in order to allow "order_cells" to be run in monocle3
  rownames(new_cds@principal_graph_aux[['UMAP']]$dp_mst) <- NULL
  colnames(SingleCellExperiment::reducedDims(new_cds)[["UMAP"]]) <- NULL
  
  ##################
  
  message("Setting all cells as belonging to one partition (multiple partitions not supported yet)")
  recreate_partition <- c(rep(1, length(new_cds@colData@rownames)))
  names(recreate_partition) <- new_cds@colData@rownames
  recreate_partition <- as.factor(recreate_partition)
  new_cds@clusters[["UMAP"]]$partitions <- recreate_partition
  
  ##################
  message("Done")
  new_cds
}

Vidya-Acadgild · 2020-04-30T16:58:40Z

@TomKellyGenetics
Thank you for sharing the codes on using Monocle3 for Seurat object. I am trying to use your script on my Seurat integrated object. I am getting an error message - could not find function "new_cell_data_set". It would be helpful to know how to resolve this. Thanks!

TomKellyGenetics · 2020-05-01T02:12:10Z

@Vidya-Acadgild Have you run library("monocle3") to import this function?

Vidya-Acadgild · 2020-05-01T02:36:09Z

@TomKellyGenetics
Thanks for the message. I re-ran the installation steps and I was able to import my Seurat object to create cds using the function new_cell_data_set (). For further analysis, I followed the steps as in the link -https://cole-trapnell-lab.github.io/monocle3/docs/clustering/
But I am stuck with this error message - "Error in sort.int(x, na.last = na.last, decreasing = decreasing, ...) : 'x' must be atomic "- when making the dot plot with the function plot_genes_by_group(). Not sure how to resolve this. I greatly appreciate any help in this regard. Thank you!

TomKellyGenetics · 2020-05-01T08:54:01Z

@Vidya-Acadgild Sorry it's unclear what issue you are having here. I suggest that you follow the steps in the vignette in order. Please note that this code is not supported by Seurat or monocle3 developers and is a workaround. If you are still having trouble, I suggest you open a new issue on more details on what you've tried already and give a minimal reproducible example. You can do this in R with dput and reprex.

Vidya-Acadgild · 2020-05-01T11:05:26Z

@TomKellyGenetics
Yes, I understand, your codes were very helpful in importing the Seurat object to Monocle3 as there was no straight forward solution available at the developers' website. Sure, I will open a new issue to discuss the follow-up errors after generating the cds with your code. Thank you!

Vidya-Acadgild · 2020-05-01T14:15:57Z

@TomKellyGenetics
A quick clarification- I imported the Seurat object (which underwent clustering and cluster specification) to monocle 3, following your code. Should I need to do the preprocessing and dimensionality reduction again as per the Monocle3 vignette - I am not sure if I understand this correctly. My apologies if I am missing something here, Thanks a lot for your time.

Wincky111 · 2020-06-05T19:27:50Z

You can also ensure that the clusters and dimension reductions match those computed in Suerat like this:
#replace monocle clusters with seurat
cds@clusters$UMAP$partitions <- sample.integrated@meta.data$seurat_clusters
names(cds@clusters$UMAP$partitions) <- rownames(sample.integrated@meta.data)
cds@clusters$UMAP$clusters <- sample.integrated@meta.data$seurat_clusters
names(cds@clusters$UMAP$clusters) <- rownames(sample.integrated@meta.data)

It wouldn't be too difficult to wrap either of these into a function but not sure on the best way to do this.

@TomKellyGenetics Hi, Tom, thank you for your codes, it is helpful.
But you did here only replace the Monocle clusters with Seurat clusters, do you know how to replace the reduce_dimension data of Monocle? Because I want the umap in Monocle is consistent with Seurat.
I tried
cds@reducedDims@ListDAta[["UMAP']] <- Data.integrated@reductions[["umap"]]@cell.embeddings
cds@reducedDims[["UMAP"]] = as.matrix(Data.integrated@reductions[["umap"]]@cell.embeddings)
reducedDims(cds) = as.matrix(molar[["pca"]]@cell.embeddings)

They didn't work. It's really puzzled me. Thanks.

TomKellyGenetics · 2020-06-08T02:34:55Z

Yes you can do that as follows.

cds@reducedDims$UMAP <-  suerat_object@reductions$umap@cell.embeddings

Note: you need to have run umap with monocle3 plot_cells(cds) first to create the objects.

Wincky111 · 2020-06-08T08:31:46Z

Yes you can do that as follows.
cds@reducedDims$UMAP <- suerat_object@reductions$umap@cell.embeddings

Note: you need to have run umap with monocle3 plot_cells(cds) first to create the objects.

Thank you for your help.
Still the same question, after using your code to create the cds object, I ran:
cds <- preprocess_cds(cds, num_dim = 20)
cds <- reduce_dimension(cds)
plot_cells(cds)
cds@reducedDims$UMAP <- seurat@reductions$umap@cell.embeddings
Still get this wrong information:
no slot of name "reducedDims" for this object of class "cell_data_set"

Is it because the version of SingleCellExperiment? Mine is 1.8.0
If I ran this:
reducedDims(cds) = as.matrix(seurat[["pca"]]@cell.embeddings)
I will get another error:
Error in vapply(value, nrow, FUN.VALUE = 0L) : values must be length 1, but FUN(X[[1]]) result is length 0

God help me==

TomKellyGenetics · 2020-06-08T23:42:09Z

Here's a minimal reproducible example.

library("Seurat")
library("monocle3")
seurat_object <- identity(pbmc_small)

library("org.Hs.eg.db")
gene_symbol <- as.list(org.Hs.egSYMBOL)
raw_count_data <- GetAssayData(seurat_object, assay = "RNA", slot = "counts")
class(raw_count_data)

cells_info <- seurat_object@meta.data

gene_name <- gene_symbol[rownames(raw_count_data)]  
gene_name <- sapply(gene_name, function(x) x[[1]][1])

#preparing cds
gene_name <- ifelse(is.na(gene_name), names(gene_name), gene_name)
gene_short_name <- gene_name
gene_id <- rownames(raw_count_data)


genes_info <- cbind(gene_id, gene_short_name)
genes_info <- as.data.frame(genes_info)
rownames(genes_info) <- rownames(raw_count_data)

cds <- new_cell_data_set(expression_data = raw_count_data,
                         cell_metadata = cells_info,
                         gene_metadata = genes_info)
cds <- preprocess_cds(cds)
cds <- reduce_dimension(cds)
cds <- cluster_cells(cds)

#replace monocle dimensions with seurat
seurat_object <- RunTSNE(seurat_object, dims = 1:2)
cds@reducedDims$TSNE <-  seurat_object@reductions$tsne@cell.embeddings
seurat_object <- RunUMAP(seurat_object, dims = 1:2)
cds@reducedDims$UMAP <-  seurat_object@reductions$umap@cell.embeddings
#replace monocle clusters with seurat
seurat_object <- FindClusters(seurat_object, dims = 1:2)
cds@clusters$UMAP$partitions <- seurat_object@meta.data$seurat_clusters
names(cds@clusters$UMAP$partitions) <- rownames(seurat_object@meta.data)
cds@clusters$UMAP$clusters <- seurat_object@meta.data$seurat_clusters
names(cds@clusters$UMAP$clusters) <- rownames(seurat_object@meta.data)


plot_cells(cds)

R version 3.6.2, Seurat_3.0.0, monocle3_0.2.0

rohanshad · 2020-06-12T19:12:57Z

I can confirm what @Wincky111 is seeing, there's no "reducedDims" slot in the cds object after preprocessing / reducing dimensions / and clustering.

This is what I see:
Error in cds@reducedDims$UMAP <- infarct@reductions$umap@cell.embeddings : no slot of name "reducedDims" for this object of class "cell_data_set"

TomKellyGenetics · 2020-06-16T01:50:42Z

@rohanshad @Wincky111 You need to run monocle3::reduce_dimension first which will create the slot, then replace it with the Seurat results.

cds <- reduce_dimension(cds)
cds@reducedDims$UMAP <- infarct@reductions$umap@cell.embeddings

This example works with monocle3 version 0.2.0.

williamhsy · 2020-06-16T23:17:57Z

Happened to meet the same error: "no slot of name "reducedDims""
If you use monocle 3 version 0.2.1, maybe you could try: reducedDims(cds)$UMAP.

It works for me!

rohanshad · 2020-06-16T23:22:19Z

Yep, above answer works!
reducedDims(cds)$UMAP <- infarct@reductions$umap@cell.embeddings replaced the monocle3 embedding with Seurat ones

brgew · 2020-06-16T23:23:25Z

Hi,

Paul Hoffman at the Satija Lab is working on tools for inter-converting Seurat and Monocle3 objects. This is a work-in-progress so it may not be feature-complete or stable but Paul encourages people to test it.

The URL is https://github.com/satijalab/seurat-wrappers/tree/feat/monocle3.

rohanshad · 2020-06-16T23:44:40Z

@williamhsy @TomKellyGenetics any luck getting the learn_graph command to work on new embeddings processed this way? use_partition = TRUE / FALSE doesn't help and 'order_cells' errors out. Hmmm...

Edit:
Was able to get everything going by following #130. Everything works as expected now.

TomKellyGenetics · 2020-06-22T02:57:02Z

As announced by Rahul Satija this is now supported by SeuratWrappers. See here for details on the code and usage.

devtools::install_github('satijalab/seurat-wrappers')
library(monocle3)
library(Seurat)
library(SeuratWrappers)

Apply it to a Seurat objects as follows:

cds <- as.cell_data_set(integrated_seurat_obj)

laijen000 · 2020-06-30T14:50:06Z

Hello! I have followed the SeuratWrappers tutorial for converting an integrated seurat v3 object to monocle3. However, when I try to plot single gene expression in pseudotime using plot_genes_in_pseudotime(), I get the error

"Error: When label_by_short_name = TRUE, rowData must have a column of gene names called gene_short_name."

rowData(cds)
DataFrame with 33538 rows and 0 columns

I wonder whether the as.cell_data_set() function to convert my integrated seurat object into a cds does not carry over gene names as a column.

Edit: I was able to get around this by adding the gene_short_name column to my rowData(cds) from a previous monocle dataframe I already had.

…_data_set" See issue in cole-trapnell-lab/monocle3#262.

yaolutian · 2020-09-13T23:58:06Z

Hello! I have followed the SeuratWrappers tutorial for converting an integrated seurat v3 object to monocle3. However, when I try to plot single gene expression in pseudotime using plot_genes_in_pseudotime(), I get the error

"Error: When label_by_short_name = TRUE, rowData must have a column of gene names called gene_short_name."

rowData(cds)
DataFrame with 33538 rows and 0 columns

I wonder whether the as.cell_data_set() function to convert my integrated seurat object into a cds does not carry over gene names as a column.

Edit: I was able to get around this by adding the gene_short_name column to my rowData(cds) from a previous monocle dataframe I already had.

Hi, I have the same problem! can you specify how you did this? thanks a lot!

deliasoto · 2020-09-23T08:17:18Z

@laijen000
Can you provide an example for adding the gene_short_name column from a previous monocle dataframe? I'm also getting:

"Error: When label_by_short_name = TRUE, rowData must have a column of gene names called gene_short_name."

or can someone else provide an example to set up the gene_short_name column after converting from a Seurat object?
Thanks!!

laijen000 · 2020-09-24T12:42:53Z

This worked for me previously!

library(monocle3)
library(SeuratWrappers)
cds <- as.cell_data_set(so)
## Calculate size factors using built-in function in monocle3
cds <- estimate_size_factors(cds)
## Add gene names into CDS
cds@rowRanges@elementMetadata@listData[["gene_short_name"]] <- rownames(so[["RNA"]])

TomKellyGenetics added the bug Something isn't working label Nov 26, 2019

hpliner added this to the v0.2.2 milestone May 13, 2020

whtns mentioned this issue May 30, 2020

'reducedDims' is not a slot in class "SingleCellExperiment" #272

Closed

hpliner modified the milestones: v0.2.2, v0.2.3 Jun 11, 2020

TomKellyGenetics closed this as completed Jun 22, 2020

jaydu1 added a commit to jaydu1/ti_monocle3 that referenced this issue Aug 14, 2020

Fix bug: no slot of name "reducedDims" for this object of class "cell…

99c8ba6

…_data_set" See issue in cole-trapnell-lab/monocle3#262.

jaydu1 mentioned this issue Aug 14, 2020

Fix bug: no slot of name "reducedDims" for this object of class "cell… dynverse/ti_monocle3#1

Open

denvercal1234GitHub mentioned this issue May 24, 2021

Monocle errors: Assays not in SingleCellExperiment object and different cluster results with cluster_cells(resolution) #507

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Requested feature: handle inputs from Seurat #262

Requested feature: handle inputs from Seurat #262

TomKellyGenetics commented Nov 26, 2019

diegoalexespi commented Dec 11, 2019

TomKellyGenetics commented Dec 23, 2019

summerrfair commented Jan 15, 2020

TomKellyGenetics commented Jan 15, 2020 •

edited

summerrfair commented Jan 16, 2020

diegoalexespi commented Feb 24, 2020 •

edited

Vidya-Acadgild commented Apr 30, 2020

TomKellyGenetics commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

TomKellyGenetics commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

Wincky111 commented Jun 5, 2020 •

edited

TomKellyGenetics commented Jun 8, 2020 •

edited

Wincky111 commented Jun 8, 2020

TomKellyGenetics commented Jun 8, 2020 •

edited

rohanshad commented Jun 12, 2020

TomKellyGenetics commented Jun 16, 2020 •

edited

williamhsy commented Jun 16, 2020

rohanshad commented Jun 16, 2020

brgew commented Jun 16, 2020

rohanshad commented Jun 16, 2020 •

edited

TomKellyGenetics commented Jun 22, 2020

laijen000 commented Jun 30, 2020 •

edited

yaolutian commented Sep 13, 2020

deliasoto commented Sep 23, 2020

laijen000 commented Sep 24, 2020

Requested feature: handle inputs from Seurat #262

Requested feature: handle inputs from Seurat #262

Comments

TomKellyGenetics commented Nov 26, 2019

diegoalexespi commented Dec 11, 2019

TomKellyGenetics commented Dec 23, 2019

summerrfair commented Jan 15, 2020

TomKellyGenetics commented Jan 15, 2020 • edited

summerrfair commented Jan 16, 2020

diegoalexespi commented Feb 24, 2020 • edited

Vidya-Acadgild commented Apr 30, 2020

TomKellyGenetics commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

TomKellyGenetics commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

Vidya-Acadgild commented May 1, 2020

Wincky111 commented Jun 5, 2020 • edited

TomKellyGenetics commented Jun 8, 2020 • edited

Wincky111 commented Jun 8, 2020

TomKellyGenetics commented Jun 8, 2020 • edited

rohanshad commented Jun 12, 2020

TomKellyGenetics commented Jun 16, 2020 • edited

williamhsy commented Jun 16, 2020

rohanshad commented Jun 16, 2020

brgew commented Jun 16, 2020

rohanshad commented Jun 16, 2020 • edited

TomKellyGenetics commented Jun 22, 2020

laijen000 commented Jun 30, 2020 • edited

yaolutian commented Sep 13, 2020

deliasoto commented Sep 23, 2020

laijen000 commented Sep 24, 2020

TomKellyGenetics commented Jan 15, 2020 •

edited

diegoalexespi commented Feb 24, 2020 •

edited

Wincky111 commented Jun 5, 2020 •

edited

TomKellyGenetics commented Jun 8, 2020 •

edited

TomKellyGenetics commented Jun 8, 2020 •

edited

TomKellyGenetics commented Jun 16, 2020 •

edited

rohanshad commented Jun 16, 2020 •

edited

laijen000 commented Jun 30, 2020 •

edited