Last updated: 2024-11-18
Checks: 7 0
Knit directory:
single-cell-jamboree/analysis/
This reproducible R Markdown analysis was created with workflowr (version 1.7.1). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.
Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.
The command set.seed(1)
was run prior to running the
code in the R Markdown file. Setting a seed ensures that any results
that rely on randomness, e.g. subsampling or permutations, are
reproducible.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.
The results in this page were generated with repository version 47d0693. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.
Note that you need to be careful to ensure that all relevant files for
the analysis have been committed to Git prior to generating the results
(you can use wflow_publish
or
wflow_git_commit
). workflowr only checks the R Markdown
file, but you know if there are other scripts or data files that it
depends on. Below is the status of the Git repository when the results
were generated:
Ignored files:
Ignored: .Rhistory
Ignored: .Rproj.user/
Untracked files:
Untracked: .DS_Store
Untracked: .gitignore
Untracked: analysis/.DS_Store
Untracked: data/.DS_Store
Untracked: data/mouse_embryo/
Untracked: output/.DS_Store
Untracked: output/mouse_embryo/
Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
These are the previous versions of the repository in which changes were
made to the R Markdown (analysis/mouse_embryo.Rmd
) and HTML
(docs/mouse_embryo.html
) files. If you’ve configured a
remote Git repository (see ?wflow_git_remote
), click on the
hyperlinks in the table below to view the files as they were in that
past version.
File | Version | Author | Date | Message |
---|---|---|---|---|
Rmd | 47d0693 | Ziang Zhang | 2024-11-18 | workflowr::wflow_publish("analysis/mouse_embryo.Rmd") |
html | 99a79f6 | Ziang Zhang | 2024-11-18 | Build site. |
Rmd | a70fb60 | Ziang Zhang | 2024-11-18 | workflowr::wflow_publish("analysis/mouse_embryo.Rmd") |
html | 070d27c | Ziang Zhang | 2024-11-15 | Build site. |
Rmd | 074ea5c | Ziang Zhang | 2024-11-15 | workflowr::wflow_publish("analysis/mouse_embryo.Rmd") |
html | 4a09688 | Ziang Zhang | 2024-11-15 | Build site. |
Rmd | 90cbbb9 | Ziang Zhang | 2024-11-15 | workflowr::wflow_publish("analysis/mouse_embryo.Rmd") |
This single cell RNA data is available here, as studied in Srivatsan et al, 2021.
The dataset considers 108725 cells and 39198 genes (after QC), measured with spatial locations in the mouse embryo.
library(Seurat)
library(Matrix)
library(data.table)
library(flashier)
library(ggplot2)
library(patchwork)
library(cowplot)
library(RColorBrewer)
library(Biobase)
library(ggpubr)
library(gridExtra)
library(fastTopics)
source('../code/plot_loadings_on_umap.R')
source("../code/Customized_Plots.R")
seurat_object <- readRDS("../data/mouse_embryo/processed_seurat/seurat_object.rds")
Y <- t(seurat_object$RNA$data)
The data contains the UMAP information that we can directly use for visualization, and compare with the “anatomical annotation” and the “cluster label” in Srivatsan et al, 2021.
umap_original_embeddings <- cbind(seurat_object$umap1, seurat_object$umap2)
p1 <- DimPlotSagnik(umap_original_embeddings, group.by = seurat_object$anatomical_annotation, pt.size = 1) +
theme(plot.title = element_text(hjust = 0.5)) +
ggtitle("UMAP by Anatomical Annotation")
p2 <- DimPlotSagnik(umap_original_embeddings, group.by = seurat_object$final_cluster_label, pt.size = 1) +
theme(plot.title = element_text(hjust = 0.5)) +
ggtitle("UMAP by Cluster Label")
# Combine plots side-by-side
combined_plot <- p1 + p2 + plot_layout(ncol = 2)
combined_plot
Version | Author | Date |
---|---|---|
4a09688 | Ziang Zhang | 2024-11-15 |
Let’s try to implement the EBMF algorithm with the
flashier
package, with different constraints and prior
distributions.
Considering the large computational cost, we will first try to run the EBMF algorithm without backfitting.
cols_to_keep <- colSums(Y != 0, na.rm = TRUE) > 0
Y <- Y[, cols_to_keep]
n <- nrow(Y)
x <- rpois(1e7, 1/n)
s1 <- sd(log(x + 1))
a <- 1
size_factors <- rowSums(Y)
size_factors <- size_factors / mean(size_factors)
shifted_log_counts <- log1p(Y / (a * size_factors))
For the first EBMF problem, we consider non-negative EBMF with
point_exponential
priors:
flashier_fit_nn <- flash(shifted_log_counts,
ebnm_fn = ebnm_point_exponential,
var_type = 2,
greedy_Kmax = 25,
S = s1,
backfit = F)
plot(flashier_fit_nn,
plot_type = "structure",
pm_which = "loadings",
pm_groups = seurat_object$final_cluster_label,
bins = 20, gap = 70)
ebnm_fn_list <- list()
ebnm_fn_list[[1]] <- flash_ebnm(
prior_family = "point_exponential"
)
ebnm_fn_list[[2]] <- flash_ebnm(
prior_family = "point_normal",
mode = "estimate"
)
flashier_fit_semi <- flash(shifted_log_counts,
ebnm_fn = ebnm_fn_list,
var_type = 2,
greedy_Kmax = 25,
S = s1,
backfit = F)
plot(flashier_fit_semi,
plot_type = "structure",
pm_which = "loadings",
pm_groups = seurat_object$final_cluster_label,
bins = 20, gap = 70)
ebnm_fn_list <- list()
ebnm_fn_list[[1]] <- flash_ebnm(
prior_family = "generalized_binary"
)
ebnm_fn_list[[2]] <- flash_ebnm(
prior_family = "point_laplace",
mode = "estimate"
)
flashier_fit_gbcd <- flash(shifted_log_counts,
ebnm_fn = ebnm_fn_list,
var_type = 2,
greedy_Kmax = 25,
S = s1,
backfit = F)
plot(flashier_fit_gbcd,
plot_type = "structure",
pm_which = "loadings",
pm_groups = seurat_object$final_cluster_label,
bins = 20, gap = 70)
res.gbcd <- fit_gbcd(Y = Y, Kmax = 25, maxiter1 = 100,
maxiter2 = 50, maxiter3 = 50,
prior = flash_ebnm(prior_family = "generalized_binary",
scale = 0.04))
saveRDS(res.gbcd, "../output/mouse_embryo/res.gbcd.rds")
Then, let’s try fitting a topic model to this dataset using
FastTopics
.
fasttopics_fit <- fit_topic_model(Y, k = 25)
Version | Author | Date |
---|---|---|
99a79f6 | Ziang Zhang | 2024-11-18 |
There are some interesting structures of the loading based on the result of EBMF and FastTopics, that add new insights to the original UMAP visualization.
First, based on the non-negative EBMF result, there are some factors that are shared across multiple cell types. For example:
The factors k10 and k3 are shared across multiple types of neurons and glial cells.
The factor k5 is shared across Chondrocytes and Connective Tissue Progenitors.
The factor k2 is shared across Erythroid Progenitors and Hepatocytes.
The factor k9 is unique to Endothelial Cells.
The factor k7 is unique to White Blood Cells.
Let’s take a look at how the UMAF plot looks if we color each cell based on its composition of these factors.
Version | Author | Date |
---|---|---|
99a79f6 | Ziang Zhang | 2024-11-18 |
The loadings from the semi-negative EBMF are more diverse, where each cell is composed of a larger number of factors. However, there are still both “unique” and “shared” factors across different cell types. For examples:
Factor k4 is shared across multiple types of neurons and glial cells.
Factor k2 is shared across Erythroid Progenitors and Hepatocytes.
Factor k3 is mostly unique to Myocytes and Cardiac Muscle.
Factor k12 is unique to Chondrocytes.
Version | Author | Date |
---|---|---|
99a79f6 | Ziang Zhang | 2024-11-18 |
The result from FastTopics looks quite similar to the semi-negative EBMF, where each cell is composed of a larger number of factors. For examples:
Factors k13 and k10 are shared across Erythroid Progenitors and Hepatocytes.
Factor k3 is unique to Hepatocytes.
Factor k8 is unique to White Blood Cells.
Version | Author | Date |
---|---|---|
99a79f6 | Ziang Zhang | 2024-11-18 |
The non-negative EBMF provides a kind of “cleaner composition” of each cell, where each cell is mostly composed of one base line factor and another factor that is ” kind of specific” to the cell type. This in a way demonstrates the advantage of the part-based representation (plus sparsity).
The semi non-negative EBMF provides a more “diverse” composition of each cell, where each cell is composed of a base line factor and multiple other factors that may appear in different cell types. The compensate of the “diversity” is the computation of this problem tends to be a bit more stable than the non-negative EBMF.
The FastTopics also provides a kind of diverse composition of each cell, where each cell is composed of multiple factors. This might be due to there is no sparsity constraint in the FastTopics model.
Overall, they all found some quite interesting structures (shared or unique) that are not obvious from the original UMAP visualization.
sessionInfo()
R version 4.3.1 (2023-06-16)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Monterey 12.7.4
Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.11.0
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
time zone: America/Chicago
tzcode source: internal
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] fastTopics_0.6-192 gridExtra_2.3 ggpubr_0.6.0
[4] Biobase_2.62.0 BiocGenerics_0.48.1 RColorBrewer_1.1-3
[7] cowplot_1.1.3 patchwork_1.3.0 ggplot2_3.5.1
[10] flashier_1.0.54 ebnm_1.1-34 data.table_1.16.2
[13] Matrix_1.6-4 Seurat_5.1.0 SeuratObject_5.0.2
[16] sp_2.1-4
loaded via a namespace (and not attached):
[1] RcppAnnoy_0.0.22 splines_4.3.1 later_1.3.2
[4] tibble_3.2.1 polyclip_1.10-7 fastDummies_1.7.4
[7] lifecycle_1.0.4 mixsqp_0.3-54 rstatix_0.7.2
[10] rprojroot_2.0.4 globals_0.16.3 lattice_0.22-6
[13] MASS_7.3-60 backports_1.5.0 magrittr_2.0.3
[16] plotly_4.10.4 sass_0.4.9 rmarkdown_2.28
[19] jquerylib_0.1.4 yaml_2.3.10 httpuv_1.6.15
[22] sctransform_0.4.1 spam_2.11-0 spatstat.sparse_3.1-0
[25] reticulate_1.39.0 pbapply_1.7-2 abind_1.4-8
[28] Rtsne_0.17 quadprog_1.5-8 purrr_1.0.2
[31] git2r_0.33.0 ggrepel_0.9.6 irlba_2.3.5.1
[34] listenv_0.9.1 spatstat.utils_3.1-0 goftest_1.2-3
[37] RSpectra_0.16-2 spatstat.random_3.3-2 fitdistrplus_1.2-1
[40] parallelly_1.38.0 leiden_0.4.3.1 codetools_0.2-20
[43] tidyselect_1.2.1 farver_2.1.2 matrixStats_1.4.1
[46] spatstat.explore_3.3-3 jsonlite_1.8.9 progressr_0.14.0
[49] Formula_1.2-5 ggridges_0.5.6 survival_3.7-0
[52] tools_4.3.1 progress_1.2.3 ica_1.0-3
[55] Rcpp_1.0.13-1 glue_1.8.0 xfun_0.48
[58] dplyr_1.1.4 withr_3.0.2 fastmap_1.2.0
[61] fansi_1.0.6 digest_0.6.37 truncnorm_1.0-9
[64] R6_2.5.1 mime_0.12 colorspace_2.1-1
[67] scattermore_1.2 gtools_3.9.5 tensor_1.5
[70] spatstat.data_3.1-2 RhpcBLASctl_0.23-42 utf8_1.2.4
[73] tidyr_1.3.1 generics_0.1.3 prettyunits_1.2.0
[76] httr_1.4.7 htmlwidgets_1.6.4 scatterplot3d_0.3-44
[79] deconvolveR_1.2-1 whisker_0.4.1 uwot_0.1.16
[82] pkgconfig_2.0.3 gtable_0.3.6 workflowr_1.7.1
[85] lmtest_0.9-40 htmltools_0.5.8.1 carData_3.0-5
[88] dotCall64_1.2 horseshoe_0.2.0 scales_1.3.0
[91] png_0.1-8 spatstat.univar_3.0-1 ashr_2.2-66
[94] knitr_1.48 rstudioapi_0.16.0 reshape2_1.4.4
[97] nlme_3.1-166 cachem_1.1.0 zoo_1.8-12
[100] Polychrome_1.5.1 stringr_1.5.1 KernSmooth_2.23-24
[103] parallel_4.3.1 miniUI_0.1.1.1 softImpute_1.4-1
[106] pillar_1.9.0 grid_4.3.1 vctrs_0.6.5
[109] RANN_2.6.2 promises_1.3.0 car_3.1-3
[112] xtable_1.8-4 cluster_2.1.6 evaluate_1.0.1
[115] invgamma_1.1 cli_3.6.3 compiler_4.3.1
[118] rlang_1.1.4 crayon_1.5.3 SQUAREM_2021.1
[121] future.apply_1.11.2 ggsignif_0.6.4 labeling_0.4.3
[124] plyr_1.8.9 fs_1.6.4 stringi_1.8.4
[127] viridisLite_0.4.2 deldir_2.0-4 munsell_0.5.1
[130] lazyeval_0.2.2 spatstat.geom_3.3-3 RcppHNSW_0.6.0
[133] hms_1.1.3 future_1.34.0 shiny_1.9.1
[136] highr_0.11 trust_0.1-8 ROCR_1.0-11
[139] igraph_2.1.1 broom_1.0.7 RcppParallel_5.1.9
[142] bslib_0.8.0