Description of scMappR R package adapted from pre-print. The
single cell mapper (scMappR) R package contains a suite of bioinformatic
tools that provide experimentally relevant cell-type specific
information to a list of differentially expressed genes (DEG). The
function scMappR_and_pathway_analysis reranks DEGs to
generate cell-type specificity scores called cell-weighted fold-changes
(cwFold-change). cwFold-changes are the original fold-changes
identifiend with bulk differntial analysis that are then scaled by the
cell-type specificity of the DEG and the cell-type proportion of the
inputted samples.
Users input a list of DEGs (gene name, fdr adjusted p-value, log2(fold-change)), read-depth normalized counts (e.g. TPM, RFPKM, CPM etc.), and a signature matrix (a gene by cell-type matrix populated with the fold-changes of cell-type specificity) into this function. Counts do not need to be log2 adjusted. scMappR then re-weights bulk DEGs by cell-type specific expression from the signature matrix, cell-type proportions from RNA-seq deconvolution and the ratio of cell-type proportions between the two conditions to account for changes gene expression driven by cell-type proportion.
With cwFold-changes calculated, scMappR uses two approaches to
utilize cwFold-changes to complete cell-type specific pathway analysis.
The two_method_pathway_enrichment re-ranks DEGs for each
cell-type based on their cwFold-change as well as the rank-order change
between cwFold-change and bulk fold-changes. The function then uses
g:Profiler to test for pathway enrichment after the DEGs were
re-ordered.
With cwFold-changes calculated, scMappR further identifies genes that
are differentially expressed in each cell-type. These differences in
expression are due to gene expression and not cell-type proportion. The
cwFoldChange_evaluate function evaluates the cell-type
specificity of cwFold-changes at the gene and cell-type levels.
At the level of the gene, cwFold-changes are scored so that they sum to 1. For each gene, cell-types whose cwFold-change are greater than the cell-type proportion while accounting for an abnormally high proportion of the fold change (+ 3 median absolute deviations from median cell-type specificity) are considered cell-type specific. At the level of the cell-type, bulk DEGs and cwFold-changes for each cell-type are correlated. The difference in the rank of DEG is also measured.
The process_dgTMatrix_lists function in the scMappR
package contains an automated scRNA-seq processing pipeline where users
input scRNA-seq count data. These data are processed, clustered, and
eventually converted into a signature matrix. scRNA-seq counts are
normalized and processed using the Seurat R package (scTransform or
Seurat V4 as options). We then identify cell-types markers and cell-type
lavels using Seurat using gene set enrichment methods Fisher’s
exact-test and GSVA using gene set markers databases from CellMarker and
PanglaoDB. Finally, custom scripts convert cell-type markers into
signature matrices.
This function can be applied to any scRNA-seq count data, however it can only label cell-types containing human or mouse gene symbols. We we apply this pipeline to the scRNA-seq datasets stored in the panglaoDB database every 3 months, allowing us to store hundreds of pre-computed signature matrices for users to apply to their own data.
The functions tissue_by_celltype_enrichment,
tissue_scMappR_internal, and
tissue_scMappR_custom combine these consistently processed
scRNAseq count data with gene-set enrichment tools to allow for
cell-type marker enrichment of a generic gene list (e.g. GWAS hits).
Sokolowski,D.J., Faykoo-Martinez,M., Erdman,L., Hou,H., Chan,C., Zhu,H., Holmes,M.M., Goldenberg,A. and Wilson,M.D. (2021) Single-cell mapper (scMappR): using scRNA-seq to infer cell-type specificities of differentially expressed genes. NAR Genomics and Bioinformatics. 3(1). Iqab011. doi 10.1093/nargab/lqab011
scMappR relies on the following dependencies which should be downlaoded/updated with scMappR automatically. Please ensure that these packages are not open when installing scMappR.
Install GSVA and pcaMethods from bioconductor first, as
devtools::install_github() will automatically install CRAN
dependencies.
```{r install_developter, eval=FALSE}
if (!requireNamespace(“BiocManager”, quietly = TRUE)) install.packages(“BiocManager”) if (!requireNamespace(“devtools”, quietly = TRUE)) install.packages(“devtools”)
if (!requireNamespace(“pcaMethods”, quietly = TRUE)) BiocManager::install(“pcaMethods”)
if (!requireNamespace(“GSVA”, quietly = TRUE)) BiocManager::install(“GSVA”)
if (!requireNamespace(“preprocessCore”, quietly = TRUE)) BiocManager::install(“preprocessCore”)
devtools::install_github(“wilsonlabgroup/scMappR”)
2. CRAN (Stable Release)
```{r install_cran, eval=FALSE}
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
if (!requireNamespace("devtools", quietly = TRUE))
install.packages("devtools")
if (!requireNamespace("pcaMethods", quietly = TRUE))
BiocManager::install("pcaMethods")
if (!requireNamespace("GSVA", quietly = TRUE))
BiocManager::install("GSVA")
# For R Version >= 4.2 individuals have reported needing
# to also install preprocessCore from bioconductor
if (!requireNamespace("preprocessCore", quietly = TRUE))
BiocManager::install("preprocessCore")
install.packages("scMappR")
Link to data used in scMappR: https://github.com/wilsonlabgroup/scMappR_Data
To run scMappR locally, please download all .rda files in this data download repository. In many of the functions, the “rda_path” argument can be changed to wherever you would like to download these files to. It however assumes “~/scMappR/data”.
If these functions do not detect these rda files, they will temporarily download them with the downloader R package; however, these rda files must already be downloaded to use the examples that are not automatically run.
```{r tissue_scMappR_internal, eval=FALSE }
signature_matrix_download <- get_signature_matrices(“all”) # get all matrices and celltype labels
## Most commonly used functions in scMappR.
Below describes the primary ways scMappR can measure cell-type specificity within a gene list as well as process scRNA-seq data. It is strongly reccomended to set `toSave = TRUE` in functions and, when appropraite, `internet = TRUE`. Otherwise scMappR will not print files/directories and many of the results will not be printed.
* `scMappR_and_pathway_analysis()`: This function generates cell-weighted Fold-changes (cwFold-changes), visualies them in a heatmap, and completes pathway enrichment of cwFold-changes and bulk gene list.
* `process_dgTMatrix_lists()`: This function takes a list of count matrices, processes them, calls cell-types, and genreates signature matrices.
* `tissue_scMappR_custom()`: This function visualizes signature matrix, clusters subsetted genes, completes enrichment of individual cell-types and co-enrichment.
* `tissue_scMappR_internal()`: This function loops through every signature matrix in a particular tissue and generates heatmaps, cell-type preferences, and co-enrichment.
* `tissue_by_celltype_enrichment()`: This function completes a fishers exact test of an input list of genes against one of the two curated tissue by cell-type marker datasets from scMappR.
### Scaling and visualizing Differentially Expressed Genes from bulk RNA-seq data
This function requires a normalized count matrix from RNA-seq (e.g. CPM, TPM, RPKM), a signature matrix such that there are fewer cell-types than there are samples, and a list of differentially expressed genes (with log2Fold-Change and adjusted P-value).
Then, it will calculate cwFold-changes before estimating if there are changes in cell-type proportion between samples. If `toSave = TRUE` then STV's will be visualized. Additionally, if `internet = TRUE`, scMappR will iteratively re-order DEGs based on their STV's and complete pathway analysis with g:ProfileR or gprofiler2.
```{r scMappR_and_pathway_analysis, eval=FALSE}
data(PBMC_scMappR)
bulk_DE_cors <- PBMC_example$bulk_DE_cors
bulk_normalized <- PBMC_example$bulk_normalized
odds_ratio_in <- PBMC_example$odds_ratio_in
case_grep <- "_female"
control_grep <- "_male"
max_proportion_change <- 10
theSpecies <- "human"
toOut <- scMappR_and_pathway_analysis(bulk_normalized, odds_ratio_in,
bulk_DE_cors, case_grep = case_grep,
control_grep = control_grep, rda_path = "",
max_proportion_change = 10, print_plots = TRUE,
plot_names = "scMappR_vignette_", theSpecies = "human",
output_directory = "scMappR_vignette_",
sig_matrix_size = 3000, up_and_downregulated = TRUE,
internet = TRUE, toSave = TRUE)
A matrix, dgGMatrix, or list of these matrices are inputted where the rows are genes and the columns are indiviudal cells. The gene names must be human or mouse gene symbols.
This function returns the signature matrix and cell-type labels. If
toSave = TRUE signature matrices, all cell-type markers,
the average expression of genes from each cell-type, and cell-type
labels from gsva are stored as files in the working directory.
Additionally, if saveSCObject = TRUE, then the Seurat
object is also saved in the working directory.
```{r process_dgTMatrix_lists, eval=FALSE}
data(sm) toProcess <- list(example = sm) tst1 <- process_dgTMatrix_lists(toProcess, name = “testPropcess”, species_name = “mouse”, naming_preference = “eye”, rda_path = “~/scMappR/data”, toSave = TRUE, saveSCObject = TRUE)
#### Generating a signature matrix with multiple scRNA-seq samples.
If there are multiple scRNA-seq runs, the `process_dgTMatrix_lists` function will integrate these data with the integration anchors feature in seurat. To complete this task, each scRNA-seq run should be a differnt element in the `dgTMatrix_list` variable.
This is covered in the "Processing scRNA-seq data with multiple scRNA-seq runs." portion of the vignette.
### Cell-type markers in a list of genes.
Input a list of human or mouse gene symbols as well as a tissue of interest to intentify and visualize enriched cell-types and cell-types co-enriched by the same genes. This process is repeated for every signature matrix (i.e. scRNA-seq study) present for the inputted tissue.
#### Provided signature matrix.
Complete the same process but with a signature matrix and gene list provided by the user. Here, gene symbols do not have to be human or mouse, the symbols in the list must match the signature.
```{r tissue_scMappR_custom, eval=FALSE}
data(POA_example)
Signature <- POA_example$POA_Rank_signature
rowname <- get_gene_symbol(Signature)
rownames(Signature) <- rowname$rowname
genes <- rownames(Signature)[1:200]
internal <- tissue_scMappR_custom(genes,Signature,output_directory = "scMappR_Test_custom", toSave = TRUE)
```{r tissue_scMappR_internal, eval=FALSE }
data(Preoptic_Area) Signature <- POA_example\(POA_Rank_signature rowname <- get_gene_symbol(Signature) rownames(Signature) <- rowname\)rowname genes <- rownames(Signature)[1:60] rda_path1 = “~/Documents/scMappR/data” internal <- tissue_scMappR_internal(genes,“mouse”,output_directory = “scMappR_Test”, tissue = “hypothalamus”,rda_path = rda_path1, toSave = TRUE)
#### Non tissue-specific
While computing custom signature matrices, there are cell-type markers across tissues and studies. These markers are stored in a gmt file format and can be used for cell-type enrichment. This may be useful if there is not a particular tissue in mind. This function inputs a list of human or mouse gene symbols.
```{r tissue_by_celltype_enrichment, eval=FALSE }
data(POA_example)
POA_generes <- POA_example$POA_generes
POA_OR_signature <- POA_example$POA_OR_signature
POA_Rank_signature <- POA_example$POA_Rank_signature
Signature <- POA_Rank_signature
rowname <- get_gene_symbol(Signature)
rownames(Signature) <- rowname$rowname
genes <- rownames(Signature)[1:100]
enriched <- tissue_by_celltype_enrichment(gene_list = genes, species = "mouse",p_thresh = 0.05, isect_size = 3)