HGG-oncohistones project [source]
- 1 Configuration
- 2 Overview
- 3 Libraries
- 4 Load metadata
- 5 Post-clustering QC of integrated data
- 6 Extract samples
- 7 Projection pipeline
- 8 Analyse projections
- 9 Cell-type projections
- 10 Gene expression levels in tumor scRNAseq
- 11 Validation of tumor cell type projections across species
- 12 Reproducibility
1 Configuration
Configuration of project directory & analysis outputs:
Show full config
source(here("rr_helpers.R"))
# Set up outputs
message("Document index: ", doc_id)## Document index: 02# Specify where to save outputs
out <- here("R-4/output", doc_id); dir.create(out, recursive = TRUE)
figout <- here("R-4/figures", doc_id); dir.create(figout, recursive = TRUE)
cache <- paste0(readLines(here("include/project_root.txt")), "R-4.1.2/", basename(here()), "/", doc_id, "/")Outputs and figures will be saved at these paths, relative to project root:
## public/R-4/output/02## public/R-4/figures/02Setting a random seed:
set.seed(100)2 Overview
Derivation of consensus cell type projections for tumors, as shown in Figure 1 and used throughout the paper.
3 Libraries
# Load libraries here
library(here)
library(tidyr)
library(dplyr)
library(ggrepel)
library(ggrastr)
library(data.table)
library(readr)
library(readxl)
library(glue)
library(tibble)
library(ggplot2)
library(GenomicRanges)
library(purrr)
library(feather)
library(cowplot)
library(Signac)
library(Seurat)
source(here("include/style.R"))
source(here("code/functions/scRNAseq.R"))
ggplot2::theme_set(theme_min())4 Load metadata
meta_sc <- data.table::fread(here("data/metadata/metadata_sc.tsv"), data.table = FALSE)5 Post-clustering QC of integrated data
For the three tumor types where we performed Harmony integration on malignant cells, we'll perform a final step of QC using the clustering in the integrated space to identify distinct clusters of neurons, and clusters of proliferating cells.
5.1 H3.3K27M
seurat_H3.3 <- get(load(here("R-4/data/integrations/H3.3K27M_malignant/output/seurat_joint.harmony.Rda")))
seurat_H3.3 %<>% summarize_cell_types.Seurat(cluster_col = "COR_ref.joint_mouse_extended") %>%
DietSeurat(counts = TRUE, dimreducs = "umap")
rm(seurat_joint_harmony)umap(seurat_H3.3, label = TRUE, legend = FALSE, point_size = 0.2)
umap(seurat_H3.3, color_by = "Type", colors = palette_type, label = FALSE, legend = FALSE, point_size = 0.2)
# feature(seurat_H3.3, "STMN2", point_size = 0.2)umap(seurat_H3.3, color_by = "S.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_H3.3, color_by = "G2M.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_H3.3, color_by = "Phase", label = FALSE, point_size = 0.2)
rm(seurat_H3.3)No distinct cluster of neurons here to remove.
5.2 H3.1/2K27M
seurat_H3.12 <- get(load(here("R-4/data/integrations/H3.12K27M_malignant/output/seurat_joint.harmony.Rda")))
seurat_H3.12 %<>% summarize_cell_types.Seurat(cluster_col = "COR_ref.joint_mouse_extended") %>%
DietSeurat(counts = TRUE, dimreducs = "umap")
rm(seurat_joint_harmony)umap(seurat_H3.12, label = TRUE, legend = FALSE, point_size = 0.2)
umap(seurat_H3.12, color_by = "Type", colors = palette_type, label = FALSE, legend = FALSE, point_size = 0.2)
feature(seurat_H3.12, "STMN2", point_size = 0.2)## @ Searching assay: RNA
seurat_neurons_remove_H3.12 <- subset(seurat_H3.12, idents = "10")
neurons_remove_H3.12 <- seurat_neurons_remove_H3.12@meta.data %>%
select(Sample, Technology, Label = Type) %>%
tibble::rownames_to_column(var = "cell.barcode") %>%
separate(cell.barcode, into = c("cellname", "drop"), sep = "_") %>%
mutate(Data = case_when(
grepl("Multiome", Technology) ~ "scMultiome",
TRUE ~ "scRNAseq"
)) %>%
select(-drop, -Technology)
# prop of cell types in the cluster
sort(table(neurons_remove_H3.12$Label)/nrow(neurons_remove_H3.12), decreasing = TRUE)##
## Neurons OPC Oligodendrocytes
## 0.67374005 0.27586207 0.01591512
## Ependymal Vascular & other Proliferating OPC
## 0.01326260 0.00795756 0.00530504
## Astrocytes RGC Glial progenitors
## 0.00530504 0.00265252 0.00000000
## Neuronal progenitors Immune Normal
## 0.00000000 0.00000000 0.00000000saveRDS(neurons_remove_H3.12, file = glue("{out}/neurons_remove_H3.12.Rds"))Here, cluster 10 expresses the neuronal marker STMN2, so we will exclude it from malignant cells.
umap(seurat_H3.12, color_by = "S.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_H3.12, color_by = "G2M.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_H3.12, color_by = "Phase", label = FALSE, point_size = 0.2)
rm(seurat_H3.12)5.3 PFA
seurat_PFA <- get(load(here("data/scRNAseq/integrations/PFA_malignant/output/seurat_joint.harmony.Rda")))
seurat_PFA %<>% summarize_cell_types.Seurat(cluster_col = "COR_ref.joint_mouse_extended") %>%
DietSeurat(counts = TRUE, dimreducs = "umap")
rm(seurat_joint_harmony)umap(seurat_PFA, label = TRUE, legend = FALSE, point_size = 0.2)
umap(seurat_PFA, color_by = "Type", colors = palette_type, label = FALSE, legend = FALSE, point_size = 0.2)
feature(seurat_PFA, "STMN2", point_size = 0.2)## @ Searching assay: RNA
seurat_neurons_remove_PFA <- subset(seurat_PFA, idents = "12")
neurons_remove_PFA <- seurat_neurons_remove_PFA@meta.data %>%
select(Sample, Technology, Label = Type) %>%
tibble::rownames_to_column(var = "cell.barcode") %>%
separate(cell.barcode, into = c("cellname", "drop"), sep = "_") %>%
mutate(Data = case_when(
grepl("Multiome", Technology) ~ "scMultiome",
TRUE ~ "scRNAseq"
)) %>%
select(-drop, -Technology)
# prop of cell types in the cluster
sort(table(neurons_remove_PFA$Label)/nrow(neurons_remove_PFA), decreasing = TRUE)##
## Ependymal Glial progenitors Neurons
## 0.699248120 0.135338346 0.110275689
## Astrocytes OPC Vascular & other
## 0.032581454 0.015037594 0.005012531
## RGC Proliferating OPC Oligodendrocytes
## 0.002506266 0.000000000 0.000000000
## Neuronal progenitors Immune Normal
## 0.000000000 0.000000000 0.000000000saveRDS(neurons_remove_PFA, file = glue("{out}/neurons_remove_PFA.Rds"))Here, a subset of cells in cluster 12 expresses the neuronal marker STMN2, so we will exclude it from malignant cells.
umap(seurat_PFA, color_by = "S.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_PFA, color_by = "G2M.Score", color_by_type = "continuous", colors = ylrd, label = FALSE, point_size = 0.2)
umap(seurat_PFA, color_by = "Phase", label = FALSE, point_size = 0.2)
rm(seurat_PFA)6 Extract samples
# get all single-cell transcriptome data from scRNAseq or scMultiome
rna_samples <- c(list.files(here("data/scRNAseq/pipeline_10X/"), full.names = TRUE))
rna_samples <- rna_samples[!grepl("Makefile", rna_samples)]
rna_seurat_paths <- file.path(rna_samples, "seurat.Rda")
all(file.exists(rna_seurat_paths))## [1] TRUEwrite_lines(rna_seurat_paths, glue("{out}/rna_samples.txt"))
multi_samples <- list.files(here("R-4/data/scMultiome/pipeline_10X_Multiome/"), full.names = TRUE)
multi_samples <- multi_samples[!grepl("Makefile", multi_samples)]
multi_seurat_paths <- file.path(multi_samples, "seurat.Rda")
all(file.exists(multi_seurat_paths))## [1] TRUEwrite_lines(multi_seurat_paths, glue("{out}/multiome_samples.txt"))
sc_samples <- c(rna_samples, multi_samples)
length(sc_samples)## [1] 47datatypes_samples <- c(rep("scRNAseq", length(rna_samples)),
rep("scMultiome", length(multi_samples)))7 Projection pipeline
Each sample was projected to the mouse brain reference atlas using an in-house Snakemake-based projection workflow which performs classification using SVM, SciBet, and Spearman correlations. The results of that workflow are loaded here.
8 Analyse projections
8.1 Load results
First, load the cell type predictions, based on multiple methods, from the projections pipeline:
projections <- map2_dfr(sc_samples, datatypes_samples, function(i, datatype) {
# message("@ ", basename(i))
id <- basename(i)
path_hydra <- gsub("from_narval", "from_hydra", i)
path_hydra <- gsub("public", "stable", path_hydra)
# load gene scores file from per-sample cNMF output
proj <- data.table::fread(file.path(path_hydra, "prediction_pipeline/Prediction_Summary.tsv"),
data.table = FALSE, header = TRUE) %>%
tibble::add_column("Sample" = basename(i), .before = 1) %>%
tibble::add_column("Data" = datatype, .after = 1)
corr <- data.table::fread(file.path(i, "correlations/ref.joint_mouse_extended/correlation.predicted_labels.tsv"),
data.table = FALSE, header = TRUE)
# sanity check
all(proj$cellnames == corr$cell)
proj$Correlation1_Prediction <- corr$celltype
return(proj)
})
dim(projections)## [1] 224238 14Next, we aggregate predictions from each method into the broader cell class as used throughout the paper:
projections_agg <- projections %>%
# Correlations performed in the first pass of cell type projection
summarize_cell_types(cluster_col = "Correlation1_Prediction") %>%
dplyr::rename(Correlation1_Type = Type) %>%
# Correlations performed in the with the projection pipeline
summarize_cell_types(cluster_col = "Correlation_Prediction") %>%
dplyr::rename(Correlation2_Type = Type) %>%
summarize_cell_types(cluster_col = "SciBet_Pred") %>%
dplyr::rename(SciBet_Type = Type) %>%
summarize_cell_types(cluster_col = "SVM_Predictions") %>%
dplyr::rename(SVM_Type = Type)8.2 Projections data-wrangling
Initally, tumor cells were projected to the mouse atlas using the full mouse atlas. However, in the prediction pipeline, the reference was downsampled for computational tractability. Let's first confirm the correlations-based predictions from the pipeline (using the downsampled reference) are consistent with the initial predictions using the same method.
NOTE: if samples are profiled by two technologies, the projections from both replicates will be combined here.
p1 <- projections_agg %>%
select(Sample, Correlation1_Prediction, Correlation_Prediction) %>%
rowwise() %>%
mutate(Correlations_equal = ifelse(Correlation1_Prediction == Correlation_Prediction, "TRUE", "FALSE")) %>%
ggplot(aes(x = Sample)) +
geom_bar(aes(fill = Correlations_equal), width = 0.5) +
scale_fill_manual(values = c("TRUE" = "gray90", "FALSE" = "red")) +
rotate_x() +
ggtitle("Concordance on granular cell types") +
ylab("# cells") +
coord_flip() +
theme(legend.position = "bottom")
p2 <- projections_agg %>%
select(Sample, Correlation1_Type, Correlation2_Type) %>%
rowwise() %>%
mutate(Correlations_equal = ifelse(Correlation1_Type == Correlation2_Type, "TRUE", "FALSE")) %>%
ggplot(aes(x = Sample)) +
geom_bar(aes(fill = Correlations_equal), width = 0.5) +
scale_fill_manual(values = c("TRUE" = "gray90", "FALSE" = "red")) +
rotate_x() +
ggtitle("Concordance on cell classes") +
ylab("# cells") +
coord_flip() +
no_legend()
plot_grid(p1, p2, align = "h", axis = "tb")
# count how many neurons predicted by correlations are supported by both versions
projections_agg %>%
select(Sample, Correlation1_Type, Correlation2_Type) %>%
rowwise() %>%
mutate(Correlations_equal = ifelse(Correlation1_Type == Correlation2_Type, "TRUE", "FALSE")) %>%
filter(Correlation1_Type == "Neurons") %>%
pull(Correlations_equal) %>%
table()## .
## FALSE TRUE
## 749 6227What we want to do is count how many methods support the predictions.
projections_agg_long <- projections_agg %>%
select(-Correlation1_Type) %>%
pivot_longer(matches("_Type"), names_to = "Method", values_to = "Prediction")
# for each cell, get most common prediction and the number of methods supporting it
projections_agg_counts <- projections_agg_long %>%
group_by(Sample, Data, cellname, Prediction) %>%
count() %>%
# peel back one layer
ungroup(Prediction) %>%
# with_ties = TRUE means that for cells with two predictions equally common will be dropped,
# both predictions will be retained
slice_max(n, n = 1, with_ties = TRUE)
# if there are cases with multiple equally common predictions, drop them and replace by "Uncertain"2
n_preds_per_cell <- projections_agg_counts %>%
group_by(Sample, Data, cellname) %>%
count()
# these cases will have 3 (a different prediction from each method)
proj_uncertain <- projections_agg_counts %>%
# right join will filter the cells in projections_agg_counts to those with
# more than one equally common prediction
right_join(n_preds_per_cell %>% filter(n == 3), by = c("Sample", "Data", "cellname")) %>%
select(-n.x, -n.y, -Prediction) %>%
mutate(Consensus_class = "Uncertain",
N_methods = 0, # Uncertain
Methods = "Uncertain") %>%
distinct(Sample, Data, cellname, Consensus_class, N_methods, Methods)
proj_certain <- projections_agg_counts %>%
right_join(n_preds_per_cell %>% filter(n == 1), by = c("Sample", "Data", "cellname")) %>%
select(-n.y) %>%
dplyr::rename(Consensus_class = Prediction,
N_methods = n.x)
# for each cell with a confident consensus prediction, get the methods that support the consensus
proj_certain2 <- projections_agg_long %>%
left_join(proj_certain, by = c("Sample", "Data", "cellname")) %>%
distinct() %>%
filter(Prediction == Consensus_class) %>%
group_by(Sample, Data, cellname, Consensus_class, N_methods) %>%
mutate(Method = gsub("_Type", "", Method)) %>%
summarize(Methods = glue_collapse(Method, sep = ","))## `summarise()` has grouped output by 'Sample', 'Data', 'cellname', 'Consensus_class'. You can override using the `.groups` argument.projections_agg_consensus <- bind_rows(proj_uncertain,
proj_certain2) %>%
arrange(Sample, Data, cellname)
# sanity check we have the same # cells/rows
nrow(projections_agg)## [1] 224238nrow(projections_agg) == nrow(projections_agg_consensus)## [1] TRUEsave(projections_agg_long, projections_agg_consensus,
file = glue("{out}/projections_aggregated_consensus.Rda"))
table(projections_agg_consensus$Consensus_class)##
## Astrocytes Ependymal Glial progenitors
## 34276 17066 8891
## Immune Neuronal progenitors Neurons
## 24290 585 5834
## Oligodendrocytes OPC Proliferating OPC
## 14809 48123 459
## RGC Uncertain Vascular & other
## 4809 58332 67648.3 Evaluate predictions across methods
In this section, we'll evaluate the consensus predictions across all cells (normal, malignant).
Which combinations of methods support each predictions?
NOTE: In this figure, all cells are independent, and each column/bar describes a specific combination of methods that support a consensus. The combinations are not subsets / unions of the others.
# what combos?
unique(projections_agg_consensus$Methods)## [1] "Correlation2,SciBet,SVM" "Correlation2,SciBet"
## [3] "Uncertain" "SciBet,SVM"
## [5] "Correlation2,SVM"# get order of frequency
methods_order <- c(setdiff(
names(sort(table(projections_agg_consensus$Methods), decreasing = TRUE)), "Uncertain"),
"Uncertain")
palette_type2 <- c(palette_type, "Uncertain" = "red")
projections_agg_consensus %>%
group_by(Methods) %>%
summarize(Prop_of_all_cells = n()/nrow(projections_agg_consensus)) %>%
arrange(desc(Prop_of_all_cells))projections_agg_consensus %>%
mutate(Methods = factor(Methods, levels = methods_order)) %>%
ggplot(aes(x = Methods)) +
geom_bar(aes(fill = Consensus_class), width = 0.5) +
scale_fill_manual(values = palette_type2) +
rotate_x() +
ggtitle("Combinations of methods supporting consensus predictions") +
ylab("# cells")
8.4 Define consensus
Finally, we define the consensus predictions as those where the prediction from the correlations method is supported by at least one other. Cells which belong to neuronal clusters identified in the Harmony-integrated data are excluded.
neurons_remove <- bind_rows(neurons_remove_H3.12, neurons_remove_PFA)
projections_agg_consensus_final <- projections_agg %>%
left_join(neurons_remove %>% mutate(Consensus_class = "Normal") %>% select(-Label), by = c("cellname", "Sample", "Data")) %>%
mutate_at(vars(matches("_Type")), as.character) %>%
mutate(
Consensus_class = case_when(
# Case 1: cells are in the neuronal clusters
Consensus_class == "Normal" & cellname %in% neurons_remove$cellname ~ "Normal",
# Case 2: prediction from correlations is supported by at least one other method
(Correlation1_Type == SciBet_Type) | (Correlation1_Type == SVM_Type) ~ Correlation1_Type,
TRUE ~ "Uncertain"
),
N_methods = case_when(
Consensus_class %in% c("Uncertain", "Normal") ~ 0,
(Correlation1_Type == SciBet_Type) & (Correlation1_Type == SVM_Type) ~ 3,
TRUE ~ 2
)) %>%
select(Sample, Data, cellname, Correlation1_Type, SciBet_Type, SVM_Type, Consensus_class, N_methods)
dim(projections_agg_consensus_final)## [1] 224238 8save(projections_agg_consensus_final, file = glue("{out}/projections_agg_consensus_final.Rda"))9 Cell-type projections
Next, let's quantify cell type projections in bar-plots, for malignant cells only. Uncertain cells, lacking a consensus projection will be shown in gray, while cells lacking a consensus and with a high G2/M phase score will be shown in a separate color.
First, we need two functions for (1) summarizing the cell-type projection at a higher level, and (2) generating a bar plot to represent cell-type proportions within each sample:
# put uncertain cells in gray
palette_type3 <- c("High G2/M" = "#e65c12", palette_type, "Uncertain" = "gray90")
load_integration <- function(path, key, prediction_filters = quo(TRUE)) {
dim <- readRDS(here(path, glue("{key}/output/dimred.harmony.Rds"))) %>% .$umap
meta1 <- readRDS(here(path, glue("{key}/output/metadata.Rds")))
if (!("COR_ref.human_fetal_thalamus3" %in% colnames(meta1))) meta1$COR_ref.human_fetal_thalamus3 <- NA
if (!("gex_barcode" %in% colnames(meta1))) meta1$gex_barcode <- NA
meta <- meta1 %>%
dplyr::rename(cellname_object = cell.barcode) %>%
mutate(cellname_object = case_when(
is.na(cellname_object) & Technology == "10X Multiome" ~ paste0(orig.ident, "_", gex_barcode),
TRUE ~ cellname_object
)) %>%
tibble::rownames_to_column(var = "cell.barcode") %>%
separate(cell.barcode, into = c("cellname", "drop"), sep = "_") %>%
dplyr::select(cellname, cellname_object, Sample, GrowthFactorReceptor, Location, Molecular,
Technology, Malignant_normal_consensus, COR_ref.human_fetal_thalamus3, G2M.Score) %>%
cbind(dim) %>%
mutate(Data = case_when(
grepl("Multiome", Technology) ~ "scMultiome",
TRUE ~ "scRNAseq"
)) %>%
# simplify growth factor mutations
mutate(GrowthFactorReceptor = case_when(
grepl("ACVR1", GrowthFactorReceptor) ~ "ACVR1",
grepl("PDGFRA", GrowthFactorReceptor) ~ "PDGFRA",
grepl("BRAF", GrowthFactorReceptor) ~ "BRAF",
TRUE ~ GrowthFactorReceptor
)) %>%
inner_join(projections_agg_consensus_final,
by = c("Sample", "Data", "cellname")) %>%
mutate(Consensus_class = case_when(
Consensus_class == "Uncertain" & G2M.Score > 0.5 ~ "High G2/M",
TRUE ~ Consensus_class
)) %>%
mutate(Location = ifelse(Location == "Thalamus", "Thal.", Location),
Consensus_class = factor(Consensus_class, levels = unique(names(palette_type3)))) %>%
filter(!!prediction_filters)
return(meta)
}
plot_umaps <- function(df, size = 0.2) {
p1 <- df %>%
ggplot(aes(x = UMAP_1, y = UMAP_2)) +
rasterize(geom_point(aes(colour = Consensus_class), size = size, alpha = 0.5), dpi = 600) +
scale_colour_manual(values = palette_type3) +
theme_min() +
theme(legend.position = "bottom") +
no_ticks()
p1
}
plot_cell_type_prop <- function(df) {
# calculate # of cells per sample
sample_n <- df %>%
select(Sample, Malignant_normal_consensus, Consensus_class, Location) %>%
filter(Malignant_normal_consensus %in% c("Malignant", "Likely malignant")) %>%
group_by(Sample) %>%
summarise(n = n())
# calculate the proportion of each major cell type among malignant cells of
# each sample
cell_type_prop <- df %>%
select(Sample, Malignant_normal_consensus, Consensus_class, Location) %>%
filter(Malignant_normal_consensus %in% c("Malignant", "Likely malignant")) %>%
group_by(Sample, Location, Consensus_class) %>%
summarise(n = n()) %>%
mutate(freq = n / sum(n))
# order by certain cell Consensus_classs
samples_ordered_epen <- cell_type_prop %>%
select(Sample, Consensus_class, freq, Location) %>%
spread(Consensus_class, freq) %>%
rowwise() %>%
mutate(Ependymal = ifelse(is.na(Ependymal), 0, Ependymal)) %>%
arrange(Location, Ependymal, Astrocytes, OPC) %>%
pull(Sample)
# plot bar chart, ordered by cell Consensus_class
p1 <- sample_n %>%
mutate(Sample = factor(Sample, levels = samples_ordered_epen)) %>%
ggplot(aes(x = Sample, y = n)) +
geom_bar(fill = "gray60", stat = "identity", width = 0.8) +
rotate_x() +
theme(panel.grid = element_blank(),
axis.ticks.x = element_blank(),
axis.text.x = element_blank()) +
xlab(NULL)
p2 <- cell_type_prop %>%
mutate(Sample = factor(Sample, levels = samples_ordered_epen)) %>%
ggplot(aes(x = Sample, y = freq)) +
geom_bar(aes(fill = Consensus_class), stat = "identity", width = 0.8) +
scale_fill_manual(values = palette_type3) +
rotate_x() +
ylab("Prop. (malignant cells)") +
theme(panel.grid = element_blank(),
axis.ticks.x = element_blank())
p3 <- df %>%
distinct(Sample, GrowthFactorReceptor) %>%
mutate(Sample = factor(Sample, levels = samples_ordered_epen)) %>%
ggplot(aes(x = Sample, y = 1)) +
geom_tile(aes(fill = GrowthFactorReceptor), colour = "white", width = 0.9, height = 0.9) +
scale_fill_manual(values = palette_gfr, na.value = "gray90", guide = guide_legend(ncol = 3)) +
theme_row() +
no_legend()
p4 <- df %>%
distinct(Sample, Location) %>%
mutate(Sample = factor(Sample, levels = samples_ordered_epen)) %>%
ggplot(aes(x = Sample, y = 1)) +
geom_tile(aes(fill = Location), colour = "white", width = 0.9, height = 0.9) +
scale_fill_manual(values = palette_location, na.value = "gray90", guide = guide_legend(ncol = 3)) +
theme_row() +
no_legend()
p_all <- plot_grid(p1, p2, p3, p4, ncol = 1, align = "v", axis = "rl",
rel_heights = c(0.15, 0.55, 0.15, 0.15))
return(p_all)
}9.0.1 H3.3K27M (malignant cells)
meta_H3.3_m <- load_integration("R-4/data/integrations", "H3.3K27M_malignant",
prediction_filters = quo(Consensus_class != "Normal"))
# stats
length(unique(meta_H3.3_m$Sample))## [1] 16table(meta_H3.3_m$Technology)##
## 10X Multiome 10X Single Cell 3' 10X Single Nuclei 3'
## 29164 19585 43658nrow(meta_H3.3_m)## [1] 92407table(meta_H3.3_m$N_methods)##
## 0 2 3
## 35172 39915 17320table(meta_H3.3_m$Consensus_class)##
## High G2/M RGC Glial progenitors
## 1612 219 1913
## OPC Proliferating OPC Oligodendrocytes
## 35395 355 8682
## Astrocytes Ependymal Neuronal progenitors
## 8727 313 39
## Neurons Immune Vascular & other
## 774 562 256
## Normal Uncertain
## 0 33560save(meta_H3.3_m, file = glue("{out}/metadata_joint_H3.3_m.Rda"))plot_umaps(meta_H3.3_m)
[figure @ public/R-4/figures/02umap_H3.3K27M_m...]
plot_cell_type_prop(meta_H3.3_m %>% filter(Consensus_class != "Uncertain"))## `summarise()` has grouped output by 'Sample', 'Location'. You can override using the `.groups` argument.
[figure @ public/R-4/figures/02bar_H3.3K27M_m...]
9.0.2 H3.3K27M thalamus (malignant cells)
meta_H3.3th_m <- load_integration("R-4/data/integrations", "H3.3K27M_thalamus_malignant",
prediction_filters = quo(Consensus_class != "Normal"))
# stats
length(unique(meta_H3.3th_m$Sample))## [1] 4table(meta_H3.3th_m$Technology)##
## 10X Multiome 10X Single Cell 3' 10X Single Nuclei 3'
## 5603 11892 22867nrow(meta_H3.3th_m)## [1] 40362table(meta_H3.3th_m$N_methods)##
## 0 2 3
## 13888 18311 8163table(meta_H3.3th_m$Consensus_class)##
## High G2/M RGC Glial progenitors
## 744 117 1178
## OPC Proliferating OPC Oligodendrocytes
## 16925 181 4846
## Astrocytes Ependymal Neuronal progenitors
## 2675 47 17
## Neurons Immune Vascular & other
## 50 339 99
## Normal Uncertain
## 0 13144save(meta_H3.3th_m, file = glue("{out}/metadata_joint_H3.3th_m.Rda"))plot_umaps(meta_H3.3th_m)
[figure @ public/R-4/figures/02umap_H3.3thK27M_m...]
plot_cell_type_prop(meta_H3.3th_m %>% filter(Consensus_class != "Uncertain"))## `summarise()` has grouped output by 'Sample', 'Location'. You can override using the `.groups` argument.
[figure @ public/R-4/figures/02bar_H3.3thK27M_m...]
9.0.3 H3.1/2K27M (malignant cells)
meta_H3.12_m <- load_integration("R-4/data/integrations", "H3.12K27M_malignant",
prediction_filters = quo(Consensus_class != "Normal"))
# stats
length(unique(meta_H3.12_m$Sample))## [1] 8table(meta_H3.12_m$Technology)##
## 10X Multiome 10X Single Nuclei 3'
## 5958 9853nrow(meta_H3.12_m)## [1] 15811table(meta_H3.12_m$N_methods)##
## 0 2 3
## 5265 7084 3462table(meta_H3.12_m$Consensus_class)##
## High G2/M RGC Glial progenitors
## 519 39 126
## OPC Proliferating OPC Oligodendrocytes
## 4722 68 772
## Astrocytes Ependymal Neuronal progenitors
## 2975 1571 10
## Neurons Immune Vascular & other
## 174 43 46
## Normal Uncertain
## 0 4746save(meta_H3.12_m, file = glue("{out}/metadata_joint_H3.12_m.Rda"))plot_umaps(meta_H3.12_m)
[figure @ public/R-4/figures/02umap_H3.12K27M_m...]
plot_cell_type_prop(meta_H3.12_m %>% filter(Consensus_class != "Uncertain"))## `summarise()` has grouped output by 'Sample', 'Location'. You can override using the `.groups` argument.
[figure @ public/R-4/figures/02bar_H3.12K27M_m...]
9.0.4 PFA-EP (malignant cells)
meta_pfa_m <- load_integration("data/scRNAseq/integrations", "PFA_malignant",
prediction_filters = quo(Consensus_class != "Normal"))
# stats
length(unique(meta_pfa_m$Sample))## [1] 5table(meta_pfa_m$Technology)##
## 10X Single Cell 3' 10X Single Nuclei 3'
## 14827 8692nrow(meta_pfa_m)## [1] 23519table(meta_pfa_m$N_methods)##
## 0 2 3
## 4221 13391 5907table(meta_pfa_m$Consensus_class)##
## High G2/M RGC Glial progenitors
## 191 535 2773
## OPC Proliferating OPC Oligodendrocytes
## 236 1 25
## Astrocytes Ependymal Neuronal progenitors
## 451 14945 3
## Neurons Immune Vascular & other
## 18 73 238
## Normal Uncertain
## 0 4030save(meta_pfa_m, file = glue("{out}/metadata_joint_PFA_m.Rda"))plot_umaps(meta_pfa_m)
[figure @ public/R-4/figures/02umap_PFA_m...]
plot_cell_type_prop(meta_pfa_m %>% filter(Consensus_class != "Uncertain"))## `summarise()` has grouped output by 'Sample', 'Location'. You can override using the `.groups` argument.
[figure @ public/R-4/figures/02bar_PFA_m...]
Clean up
rm(meta_H3.3_m)
rm(meta_H3.12_m)
rm(meta_pfa_m)10 Gene expression levels in tumor scRNAseq
10.1 Load data
Load joint object of malignant cells:
seurat_h31 <- get(load(here("R-4/data/integrations/H3.12K27M_malignant/output/seurat_joint.Rda")))
rm(seurat_joint)
seurat_h33 <- get(load(here("R-4/data/integrations/H3.3K27M_malignant/output/seurat_joint.Rda")))
rm(seurat_joint)
seurat_h31_small <- DietSeurat(seurat_h31, data = TRUE, counts = FALSE)
seurat_h33_small <- DietSeurat(seurat_h33, data = TRUE, counts = FALSE)
rm(seurat_h31)
rm(seurat_h33)10.2 Expression
A function to plot levels of each gene per sample:
expr_data <- bind_rows(FetchData(seurat_h31_small,
vars = c("NKX6-1", "PAX3", "Sample",
"Molecular", "Location", "GrowthFactorReceptor", "Technology")),
FetchData(seurat_h33_small,
vars = c("NKX6-1", "PAX3", "Sample",
"Molecular", "Location", "GrowthFactorReceptor", "Technology")))
# restrict cells to those with confident predictions
expr_data_anno <- expr_data %>%
tibble::rownames_to_column(var = "cell.barcode") %>%
separate(cell.barcode, into = c("cellname", "drop"), sep = "_") %>%
dplyr::select(cellname, Sample, GrowthFactorReceptor, Location, Molecular,
Technology, `NKX6-1`, PAX3) %>%
mutate(Data = case_when(
grepl("Multiome", Technology) ~ "scMultiome",
TRUE ~ "scRNAseq"
)) %>%
inner_join(projections_agg_consensus_final,
by = c("Sample", "Data", "cellname")) %>%
# filter to cells with projections supported by confident combos
filter(!(Consensus_class %in% c("Normal", "Uncertain"))) %>%
select(Sample = Sample, Consensus_class, everything())
dim(expr_data_anno)## [1] 67781 14table(expr_data_anno$Consensus_class)##
## Astrocytes Ependymal Glial progenitors
## 11702 1884 2039
## Immune Neuronal progenitors Neurons
## 605 49 948
## Oligodendrocytes OPC Proliferating OPC
## 9454 40117 423
## RGC Vascular & other
## 258 302saveRDS(expr_data_anno, file = glue("{out}/tumor_scRNA_expr_data.Rds"))
# split by tumor type
celltype_freq <- expr_data_anno %>%
mutate(Molecular = ifelse(Molecular == "H3.2K27M", "H3.1K27M", Molecular)) %>%
group_by(Molecular) %>%
mutate(N_per_molecular = n()) %>%
group_by(Molecular, N_per_molecular, Consensus_class) %>%
count() %>%
mutate(Freq_per_molecular = n/N_per_molecular) %>%
ungroup() %>%
select(Molecular, Consensus_class, N = n, Freq = Freq_per_molecular) %>%
pivot_wider(names_from = Molecular, values_from = c(N, Freq))
celltype_freq# filter out cells which represent less than 5% of the dataset
(celltype_keep_h31 <- celltype_freq %>% filter(Freq_H3.1K27M > 0.05) %>% pull(Consensus_class))## [1] "Astrocytes" "Ependymal" "Oligodendrocytes" "OPC"(celltype_keep_h33 <- celltype_freq %>% filter(Freq_H3.3K27M > 0.05) %>% pull(Consensus_class))## [1] "Astrocytes" "Oligodendrocytes" "OPC"pct1_joint_sample <- expr_data_anno %>%
group_by(Molecular, Sample, Consensus_class) %>%
# calculate the number of cells of each cell type per sample
mutate(n_total = n()) %>%
ungroup() %>%
gather(Gene, Expression, `NKX6-1`, PAX3) %>%
group_by(n_total, Molecular, GrowthFactorReceptor, Location, Sample, Consensus_class, Gene) %>%
# count the number of cells of each cell type where each gene is detected
summarize(N_detected = sum(Expression > 0)) %>%
# divide by the total # of cells in the cell type
mutate(pct1 = round(N_detected / n_total, 2)) %>%
ungroup() %>%
select(-n_total) %>%
mutate(Consensus_class = factor(Consensus_class, levels = rev(names(palette_type))))## `summarise()` has grouped output by 'n_total', 'Molecular', 'GrowthFactorReceptor', 'Location', 'Sample', 'Consensus_class'. You can override using the `.groups` argument.plot_boxplot_per_sample <- function(df, title, color, label_ACVR1 = FALSE) {
if (label_ACVR1) df <- df %>%
mutate(ACVR1_simple = case_when(
grepl("ACVR1-", GrowthFactorReceptor) ~ "Mutant",
TRUE ~ "WT"
))
p1 <- df %>%
mutate(Gene = factor(Gene, levels = c("NKX6-1", "PAX3"))) %>%
complete(Gene, nesting(Consensus_class), fill = list(pct1 = 0)) %>%
ggplot(aes(x = Consensus_class, y = pct1)) +
geom_boxplot(lwd = 0.25, fill = color, outlier.shape = NA)
if (label_ACVR1) {
p1 <- p1 +
geom_point(aes(shape = ACVR1_simple), stat = "identity", alpha = 0.7) +
scale_shape_manual(values = palette_acvr1_simple)
} else p1 <- p1 + geom_point(stat = "identity", colour = "black", alpha = 0.7)
p1 <- p1 +
scale_alpha(guide = FALSE) +
facet_wrap(~ Gene, nrow = 1) +
coord_flip() +
rotate_x() +
# no_legend() +
ylab("Proportion of cells in which the gene is detected") +
ylim(c(0, 1)) +
ggtitle(title) +
theme(legend.position = "bottom")
return(p1)
}Encoding ACVR1 status:
p1 <- plot_boxplot_per_sample(pct1_joint_sample %>%
filter(Molecular %in% c("H3.1K27M", "H3.2K27M") &
Location == "Pons" &
Consensus_class %in% celltype_keep_h31),
title = "H3.1/2K27M Pons",
"orange",
label_ACVR1 = TRUE)
p2 <- plot_boxplot_per_sample(pct1_joint_sample %>%
filter(Molecular == "H3.3K27M" &
Location == "Pons" &
Consensus_class %in% celltype_keep_h33),
title = "H3.3K27M Pons",
"red3",
label_ACVR1 = TRUE)
plot_grid(p1, p2, ncol = 1, align = "h", axis = "tb")
11 Validation of tumor cell type projections across species
11.1 Load human fetal thalamus data
hf_thalamus_annotation <- data.table::fread(here("output/01A/hf_thalamus_annotation.tsv"),
data.table = FALSE)
summarize_cell_types_bhaduri <- function(df, cluster_col) {
cc_quo <- rlang::sym(cluster_col)
df %>%
mutate(
# Define some broader cell type classes
Type = case_when(
grepl("GLIP|OPAS|Glial", !!cc_quo) ~ "Glial progenitors",
grepl("RG|[Rr]adial|NSC|prog|Dividing", !!cc_quo) ~ "RGC",
grepl("EXIP|INIP|NEURP|IP", !!cc_quo) & !grepl("VIP", !!cc_quo) ~ "Neuronal progenitors",
grepl("MGE|CGE|LGE|SST|PV|INH|CEX|PEX|GABAN|NEU|SPN|NRGN|UBCN|MFN|SERN|[Nn]euron", !!cc_quo) ~ "Neurons",
grepl("OPC-P|Proliferating OPC", !!cc_quo) ~ "Proliferating OPC",
grepl("OPC", !!cc_quo) ~ "OPC",
grepl("NFOL|MOL|Oligo", !!cc_quo) ~ "Oligodendrocytes",
grepl("ASTR|Astro", !!cc_quo) ~ "Astrocytes",
grepl("EPEN|ASEP|Epen", !!cc_quo) ~ "Ependymal",
grepl("MGL|MAC|Micro", !!cc_quo) ~ "Immune",
grepl("T|Fibr|Mur|Micro|Mixed|UNR|Other|ENDO|PERI|MNG|Endo", !!cc_quo) ~ "Vascular & other",
TRUE ~ "Vascular & other")) %>%
mutate(Type = factor(Type, levels = c("RGC",
"Glial progenitors",
"Proliferating OPC",
"OPC",
"Oligodendrocytes",
"Astrocytes",
"Ependymal",
"Neuronal progenitors",
"Neurons",
"Immune",
"Vascular & other",
"Normal")))
}
hf_thalamus_info_clusters <- data.table::fread(here("output/01A/info_clusters3.tsv"),
data.table = FALSE) %>%
filter(!grepl("EXCLUDE", ID_20220321)) %>%
summarize_cell_types_bhaduri("ID_20220321") %>%
mutate(Label = paste0(Sample, " #", Cluster_number))11.2 Load tumor data
We'll load the combined metadata for all H3.3K27M thalamic samples, restricted to malignant cells:
meta_H3.3thal_m <- load_integration("R-4/data/integrations", "H3.3K27M_thalamus_malignant",
prediction_filters = quo(!(Consensus_class %in% c("Normal", "Uncertain"))))
length(unique(meta_H3.3thal_m$Sample))## [1] 4table(meta_H3.3thal_m$Technology)##
## 10X Multiome 10X Single Cell 3' 10X Single Nuclei 3'
## 4483 7760 14975nrow(meta_H3.3thal_m)## [1] 27218table(meta_H3.3thal_m$N_methods)##
## 0 2 3
## 744 18311 8163table(meta_H3.3thal_m$Consensus_class)##
## High G2/M RGC Glial progenitors
## 744 117 1178
## OPC Proliferating OPC Oligodendrocytes
## 16925 181 4846
## Astrocytes Ependymal Neuronal progenitors
## 2675 47 17
## Neurons Immune Vascular & other
## 50 339 99
## Normal Uncertain
## 0 0meta_H3.3thal_m <- meta_H3.3thal_m %>%
left_join(hf_thalamus_info_clusters %>% select(Cluster, Type), by = c("COR_ref.human_fetal_thalamus3" = "Cluster")) %>%
dplyr::rename(Type_human = Type)
# total agreement
meta_H3.3thal_m %>%
mutate(Type_human = as.character(Type_human),
Consensus_class = as.character(Consensus_class)) %>%
filter(Type_human == Consensus_class) %>% { nrow(.)/nrow(meta_H3.3thal_m) }## [1] 0.805680111.3 Mouse vs. human projections
UMAP coloured by human projections:
meta_H3.3thal_m %>%
ggplot(aes(x = UMAP_1, y = UMAP_2)) +
rasterize(geom_point(aes(colour = Type_human), size = 0.2, alpha = 0.5), dpi = 600) +
scale_colour_manual(values = palette_type3) +
theme_min() +
theme(legend.position = "bottom") +
no_ticks()
[figure @ public/R-4/figures/02umap_H3.3thK27M_m_labelled_thalamus...]
Plot the agreement as a confusion matrix representing the proportion of each mouse cell types that are predicted to human cell types:
# encode number of cells as well as proportion of agreement
p3 <- meta_H3.3thal_m %>%
group_by(Consensus_class) %>%
mutate(Total_per_class_mouse = n()) %>%
group_by(Consensus_class, Total_per_class_mouse, Type_human) %>%
count() %>%
mutate(Prop = n/Total_per_class_mouse) %>%
mutate(Type_human = factor(Type_human, levels = setdiff(names(palette_type), "Normal")),
Consensus_class = factor(Consensus_class, levels = rev(setdiff(names(palette_type), "Normal")))) %>%
ungroup() %>%
complete(Consensus_class, Type_human, fill = list(Prop = 0)) %>%
filter(Consensus_class != "Uncertain" & Type_human != "Uncertain") %>%
ggplot(aes(x = Type_human, y = Consensus_class)) +
geom_point(aes(colour = Prop, size = n)) +
scale_colour_gradientn(colors = colorRampPalette(c("gray90", "red"))(n=100), limits = c(0, 1)) +
scale_size_area(breaks = seq(2000, 10000, by = 2000), max_size = 10) +
geom_text(aes(label = round(Prop, 2)), colour = "black", size = 3) +
scale_x_discrete(position = "top") +
theme_min() +
theme(panel.border = element_blank()) +
rotate_x() +
theme(axis.text.x = element_text(hjust = 0)) +
ggtitle("Proportion of cells (labelled by mouse) \nprojected to each human cell class") +
xlab("Human") + ylab("Mouse")
p4 <- meta_H3.3thal_m %>%
group_by(Type_human) %>%
mutate(Total_per_class_human = n()) %>%
group_by(Type_human, Total_per_class_human, Consensus_class) %>%
count() %>%
mutate(Prop = n/Total_per_class_human) %>%
mutate(Type_human = factor(Type_human, levels = rev(setdiff(names(palette_type), "Normal"))),
Consensus_class = factor(Consensus_class, levels = setdiff(names(palette_type), "Normal"))) %>%
ungroup() %>%
complete(Consensus_class, Type_human, fill = list(Prop = 0)) %>%
filter(Consensus_class != "Uncertain" & Type_human != "Uncertain") %>%
ggplot(aes(x = Consensus_class, y = Type_human)) +
geom_point(aes(colour = Prop, size = n)) +
scale_colour_gradientn(colors = colorRampPalette(c("gray90", "red"))(n=100), limits = c(0, 1)) +
scale_size_area(breaks = seq(2000, 10000, by = 2000), max_size = 10) +
geom_text(aes(label = round(Prop, 2)), colour = "black", size = 3) +
scale_x_discrete(position = "top") +
theme_min() +
theme(panel.border = element_blank()) +
rotate_x() +
theme(axis.text.x = element_text(hjust = 0)) +
ggtitle("Proportion of cells (labelled by human) \nprojected to each mouse cell class") +
xlab("Mouse") + ylab("Human")
plot_grid(p3, p4, ncol = 2)
12 Reproducibility
This document was last rendered on:
## 2022-06-29 11:08:00The git repository and last commit:
## Local: master /lustre06/project/6004736/sjessa/from_narval/HGG-oncohistones/public
## Remote: master @ origin (git@github.com:fungenomics/HGG-oncohistones.git)
## Head: [6e4c415] 2022-06-29: Add functions for R 3.6The random seed was set with set.seed(100)
The R session info:
## R version 4.1.2 (2021-11-01)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Rocky Linux 8.5 (Green Obsidian)
##
## Matrix products: default
## BLAS/LAPACK: /cvmfs/soft.computecanada.ca/easybuild/software/2020/Core/flexiblas/3.0.4/lib64/libflexiblas.so.3.0
##
## locale:
## [1] LC_CTYPE=en_CA.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_CA.UTF-8 LC_COLLATE=en_CA.UTF-8
## [5] LC_MONETARY=en_CA.UTF-8 LC_MESSAGES=en_CA.UTF-8
## [7] LC_PAPER=en_CA.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_CA.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] parallel stats4 stats graphics grDevices datasets utils
## [8] methods base
##
## other attached packages:
## [1] magrittr_2.0.1 viridis_0.5.1 viridisLite_0.3.0
## [4] RColorBrewer_1.1-2 SeuratObject_4.0.4 Seurat_4.0.0
## [7] Signac_1.3.0 cowplot_1.1.1 feather_0.3.5
## [10] purrr_0.3.4 GenomicRanges_1.42.0 GenomeInfoDb_1.26.7
## [13] IRanges_2.24.1 S4Vectors_0.28.1 BiocGenerics_0.36.1
## [16] tibble_3.1.6 glue_1.6.1 readxl_1.3.1
## [19] readr_2.1.1 data.table_1.14.2 ggrastr_0.2.3
## [22] ggrepel_0.9.1 ggplot2_3.3.5 dplyr_1.0.7
## [25] tidyr_1.1.4 here_1.0.1
##
## loaded via a namespace (and not attached):
## [1] fastmatch_1.1-3 plyr_1.8.6 igraph_1.2.11
## [4] lazyeval_0.2.2 splines_4.1.2 BiocParallel_1.24.1
## [7] listenv_0.8.0 SnowballC_0.7.0 scattermore_0.7
## [10] digest_0.6.29 htmltools_0.5.2 fansi_1.0.2
## [13] tensor_1.5 cluster_2.1.2 ROCR_1.0-11
## [16] tzdb_0.2.0 globals_0.14.0 Biostrings_2.58.0
## [19] matrixStats_0.61.0 docopt_0.7.1 colorspace_2.0-2
## [22] xfun_0.29 sparsesvd_0.2 crayon_1.4.2
## [25] RCurl_1.98-1.5 jsonlite_1.7.3 spatstat_1.64-1
## [28] spatstat.data_2.1-2 survival_3.2-13 zoo_1.8-9
## [31] polyclip_1.10-0 gtable_0.3.0 zlibbioc_1.36.0
## [34] XVector_0.30.0 leiden_0.3.9 future.apply_1.8.1
## [37] abind_1.4-5 scales_1.1.1 DBI_1.1.2
## [40] miniUI_0.1.1.1 Rcpp_1.0.8 xtable_1.8-4
## [43] reticulate_1.23 htmlwidgets_1.5.4 httr_1.4.2
## [46] ellipsis_0.3.2 ica_1.0-2 farver_2.1.0
## [49] pkgconfig_2.0.3 ggseqlogo_0.1 sass_0.4.0
## [52] uwot_0.1.11 deldir_1.0-6 utf8_1.2.2
## [55] labeling_0.4.2 tidyselect_1.1.1 rlang_0.4.12
## [58] reshape2_1.4.4 later_1.3.0 munsell_0.5.0
## [61] cellranger_1.1.0 tools_4.1.2 generics_0.1.1
## [64] ggridges_0.5.3 evaluate_0.14 stringr_1.4.0
## [67] fastmap_1.1.0 yaml_2.2.1 goftest_1.2-3
## [70] knitr_1.37 fitdistrplus_1.1-6 RANN_2.6.1
## [73] pbapply_1.5-0 future_1.23.0 nlme_3.1-153
## [76] mime_0.12 slam_0.1-50 RcppRoll_0.3.0
## [79] compiler_4.1.2 beeswarm_0.4.0 plotly_4.10.0
## [82] png_0.1-7 spatstat.utils_2.3-0 tweenr_1.0.2
## [85] bslib_0.3.1 stringi_1.7.6 highr_0.9
## [88] lattice_0.20-45 Matrix_1.3-4 vctrs_0.3.8
## [91] pillar_1.6.4 lifecycle_1.0.1 BiocManager_1.30.15
## [94] lmtest_0.9-39 jquerylib_0.1.4 RcppAnnoy_0.0.19
## [97] bitops_1.0-7 irlba_2.3.5 httpuv_1.6.5
## [100] patchwork_1.1.1 R6_2.5.1 promises_1.2.0.1
## [103] renv_0.15.5 lsa_0.73.2 KernSmooth_2.23-20
## [106] gridExtra_2.3 vipor_0.4.5 parallelly_1.30.0
## [109] codetools_0.2-18 MASS_7.3-54 assertthat_0.2.1
## [112] rprojroot_2.0.2 withr_2.4.3 qlcMatrix_0.9.7
## [115] sctransform_0.3.3 Rsamtools_2.6.0 GenomeInfoDbData_1.2.4
## [118] mgcv_1.8-38 hms_1.1.1 grid_4.1.2
## [121] rpart_4.1-15 rmarkdown_2.11 Cairo_1.5-14
## [124] Rtsne_0.15 git2r_0.29.0 ggforce_0.3.3
## [127] shiny_1.7.1 ggbeeswarm_0.6.0The resources requested when this document was last rendered:
## #SBATCH --time=00:20:00
## #SBATCH --cpus-per-task=1
## #SBATCH --mem=60G
A project of the Kleinman Lab at McGill University, using the rr reproducible research template.