5 Single-Cell Analysis

Most of these code are reliant on Seurat. You may now and then have to use scanpy in python, or revert to the SingleCellExperiment object utilized in scater.

5.1 Useful links and resources

Fine-tuning UMAP visualizations https://jlmelville.github.io/uwot/abparams.html
TSNE animation https://distill.pub/2016/misread-tsne/
Simple explanantion on TSNE https://www.cancer.gov/about-nci/organization/ccg/blog/2020/interview-t-sne
Different integration methods https://swaruplab.bio.uci.edu/tutorial/integration/integration_tutorial.html
Illustration of how UMAPs can be misleading https://pair-code.github.io/understanding-umap/ https://blog.bioturing.com/2022/01/14/umap-vs-t-sne-single-cell-rna-seq-data-visualization/

Single cell best practices from the Theis lab https://www.sc-best-practices.org/preamble.html https://broadinstitute.github.io/2019_scWorkshop/index.html

Deep explanation of Seurat’s AddModuleScore function https://www.waltermuskovic.com/2021/04/15/seurat-s-addmodulescore-function/

5.1.1 Psudotime analysis

https://broadinstitute.github.io/2019_scWorkshop/functional-pseudotime-analysis.html
List of single cell pseudotime packages publishe#d https://github.com/agitter/single-cell-pseudotime

5.1.2 Spatial Transciptomics

https://scimap.xyz/tutorials/md/spatial_biology_scimap/

5.1.3 Downloading data

bam files uploaded to SRA will be missing certain 10x flags and will not work. You need to get the original bam files via SRA Data Access Tab and downloaded via wget and run bamtofastq.
for 10x, need to use –split-filles and –include-technical.

5.1.4 CellRanger

5.2 Seurat

5.2.1 Integration

applying sctransform to each sample in the list
selecting integration features,
finding integration anchors,
integrate data at this point, DefaultAssay is integrated, and there is an associated VariableFeatures, which really is output of SelectIntegrationFetures. rownames(integrated[[“SCT”]](scale.data?)) is the same as SelectIntegration Features again.
https://github.com/satijalab/seurat/issues/6185 explanation of the different variable gene options from sct integrated models

After merge() variable features that have been calculated by SCTransform is reset.

5.2.2 Plotting

Use scCustomize for improved plotting over Seurat’s defaults.

To plot heatmaps with extra metadata information, use the scillius package.

To get statistics on differential gene levels, use the package ggbetweenstats from ggstatsplot.

cell_expression <- rna_integrated@assays$RNA@data[c("MIPOL1", "ETV1", "DGKB", "FOXA2", "AR", "SYP"),] %>% as.matrix() %>% t()
meta_data <- rna_integrated@meta.data
plot_data <- cbind(meta_data, cell_expression)
saveRDS(plot_data, "scRNAseq_violin.RDS")

5.2.3 Utility functions

Renaming all identities

cell_types <- c("0" = "", "1" = "", "2" = "", "3" = "", "4" = "", "5" = "", "6" = "")
Idents(merged) <- "SCT_snn_res.0.2"
merged <- RenameIdents(merged, cell_types)
merged$status <- Idents(merged)
DimPlot_scCustom(integrated, group.by = "cell_types", colors_use = polychrome_pal)
merged$status <- factor(merged$status, 
                        levels = c("")) # useful for rearranging levels
Idents(merged) <- "status"

Renaming a few identities

new_groups <- case_when(seu_obj$clust == "a" ~ "Tumor",
          .default = seu_obj$clust) %>% as.factor()
names(new_groups) <- names(seu_obj$clust)
seu_obj$new_groups <- new_groups
seu_obj$new_groups <- factor(seu_obj$new_groups, 
                        levels = c("")) # have to do this again

Adding new clusters, cluster 13 was not present previously

seu_obj$manual_clusters <- seu_obj$integrated_snn_res.0.6
levels(seu_obj$manual_clusters) <- c(levels(seu_obj$manual_clusters), "13")
seu_obj$manual_clusters[names(seu_obj$manual_clusters) %in% select_cells] <- "13"

--- execute: eval: false format: html: link-external-newwindow: true --- # Single-Cell Analysis Most of these code are reliant on Seurat. You may now and then have to use scanpy in python, or revert to the SingleCellExperiment object utilized in scater. ## Useful links and resources Fine-tuning UMAP visualizations <https://jlmelville.github.io/uwot/abparams.html>\ TSNE animation <https://distill.pub/2016/misread-tsne/>\ Simple explanantion on TSNE <https://www.cancer.gov/about-nci/organization/ccg/blog/2020/interview-t-sne>\ Different integration methods <https://swaruplab.bio.uci.edu/tutorial/integration/integration_tutorial.html>\ Illustration of how UMAPs can be misleading <https://pair-code.github.io/understanding-umap/> <https://blog.bioturing.com/2022/01/14/umap-vs-t-sne-single-cell-rna-seq-data-visualization/> Single cell best practices from the Theis lab <https://www.sc-best-practices.org/preamble.html> <https://broadinstitute.github.io/2019_scWorkshop/index.html> Deep explanation of Seurat's AddModuleScore function <https://www.waltermuskovic.com/2021/04/15/seurat-s-addmodulescore-function/> ### Psudotime analysis <https://broadinstitute.github.io/2019_scWorkshop/functional-pseudotime-analysis.html>\ List of single cell pseudotime packages publishe#d <https://github.com/agitter/single-cell-pseudotime> ### Spatial Transciptomics <https://scimap.xyz/tutorials/md/spatial_biology_scimap/> ### Downloading data bam files uploaded to SRA will be missing certain 10x flags and will not work. You need to get the original bam files via SRA Data Access Tab and downloaded via wget and run bamtofastq.\ for 10x, need to use --split-filles and --include-technical. ### CellRanger ## Seurat ### Integration - applying sctransform to each sample in the list - selecting integration features, - finding integration anchors, - integrate data at this point, DefaultAssay is integrated, and there is an associated VariableFeatures, which really is output of SelectIntegrationFetures. rownames(integrated\[\["SCT"\]\]@scale.data) is the same as SelectIntegration Features again.\ https://github.com/satijalab/seurat/issues/6185 explanation of the different variable gene options from sct integrated models After `merge()` variable features that have been calculated by SCTransform is reset. ### Plotting Use [scCustomize](https://samuel-marsh.github.io/scCustomize/index.html) for improved plotting over Seurat's defaults. To plot heatmaps with extra metadata information, use the [scillius](https://scillus.netlify.app/) package. To get statistics on differential gene levels, use the package `ggbetweenstats` from [ggstatsplot](https://indrajeetpatil.github.io/ggstatsplot/index.html). ```{r} cell_expression <- rna_integrated@assays$RNA@data[c("MIPOL1", "ETV1", "DGKB", "FOXA2", "AR", "SYP"),] %>% as.matrix() %>% t() meta_data <- rna_integrated@meta.data plot_data <- cbind(meta_data, cell_expression) saveRDS(plot_data, "scRNAseq_violin.RDS") ``` ### Utility functions Renaming all identities ```{r} cell_types <- c("0" = "", "1" = "", "2" = "", "3" = "", "4" = "", "5" = "", "6" = "") Idents(merged) <- "SCT_snn_res.0.2" merged <- RenameIdents(merged, cell_types) merged$status <- Idents(merged) DimPlot_scCustom(integrated, group.by = "cell_types", colors_use = polychrome_pal) merged$status <- factor(merged$status, levels = c("")) # useful for rearranging levels Idents(merged) <- "status" ``` Renaming a few identities ```{r} new_groups <- case_when(seu_obj$clust == "a" ~ "Tumor", .default = seu_obj$clust) %>% as.factor() names(new_groups) <- names(seu_obj$clust) seu_obj$new_groups <- new_groups seu_obj$new_groups <- factor(seu_obj$new_groups, levels = c("")) # have to do this again ``` Adding new clusters, cluster 13 was not present previously ```{r} seu_obj$manual_clusters <- seu_obj$integrated_snn_res.0.6 levels(seu_obj$manual_clusters) <- c(levels(seu_obj$manual_clusters), "13") seu_obj$manual_clusters[names(seu_obj$manual_clusters) %in% select_cells] <- "13" ```