A single embryo, single cell time-resolved model for mouse gastrulation

This repository contains all the code for reproducing the analysis from the gastrulation flow paper Mittnenzweig et al. (2021). The analysis is done with the metacell R package, that also contains the code for generating the network flow model.

Quick links

• Metacell paper: Baran et al. 2019 Genome Biology

• Metacell R package

• Raw FASTQ files and processed UMI tables are available under GEO accession GSE169210

Requirements (R packages)

• metacell
• lpsymphony
• pheatmap
• gridExtra
• Matrix
• tidyverse
• shape
• umap
• qlcMatrix
• ggrepel

Usage

After cloning the github repository, users should open an R session in the repository root directory and download/initialize the scRNA database (~ 4.7 GB):

# Loading code and downloading required data files
source("scripts/download_data.r")

The repository root directory should now contain the subfolders scripts/ containing all the R scripts, scrna_db containing the metacell R objects, config/, data/ containing additional data generated by the scripts and figs/paper_figs/.

Generating all figures

All figures of the paper can be regenerated by running:

# load all scripts
source("scripts/initialize_scripts.r")
generate_all_figures()

Please note that this will take some time. If you are interested in regenerating a specific figure, see the paragraph below.

Regenerating plots for a specific figure

For each figure (Figures 1-7 and S1-7), there is a corresponding script in scripts/generate_paper_figures/. Each script contains a function gen_fig_xyz_plots() at the top, that contains further subfunctions and explanations related to the analysis of that figure. E.g., for regenerating the plots of figure 1, users should run the following code:

# load metacell package
library("metacell")
# initializing the metacell scrna database
scdb_init("scrna_db")

# Generating plots of Figure 1
source("scripts/generate_paper_figures/fig_1.r")
gen_fig_1_plots()

The content of gen_fig_1_plots() looks as follwos:

gen_fig_1_plots = function() {
  
  if(!dir.exists("figs/paper_figs")) {
    dir.create("figs/paper_figs")
  }
  dir_name = "figs/paper_figs/fig1"
  
  if(!dir.exists(dir_name)) {
    dir.create(dir_name)
  }
  
  fig1_b()
  fig1_cde()
  fig1_f()
  fig1_g_mc_time_distributions()
  fig1_g_heatmap(plot_pdf = T)
  fig1_h()
  
}

Figure plots are saved in figs/paper_figs/fig1/

Computing metacell object, manifold graph and network flow model for wildtype manifold

Standard metacell analysis is performed as described in Baran et al. 2019. To recompute the metacell object, please run

source("scripts/generate_mc_mgraph_network/gen_mc.r")

This will generate a metacell object with id sing_emb_wt10_bs500f. Note, that because of random seeding of the boostrap procedure involved in calculating the metacell cover, the computed metacell cover will slightly deviate from sing_emb_wt10_recolored used in the paper. Manifold graphs and 2D projections can be recomputed through

source("scripts/generate_mc_mgraph_network/gen_mgraph.r")
source("scripts/generate_mc_mgraph_network/gen_mgraph_umap.r")
generate_mgraph_wt10()
gen_mc2d_umap_wt10()

The network flow model can be generated using

source("scripts/generate_mc_mgraph_network/gen_network.r")
build_sing_emb_wt10_network()

Metacells were clustered and annotated using the network flow model.

source("scripts/generate_mc_mgraph_network/annot_mc_by_flows.r")
cluster_metacells_by_flow(mct_id = "sing_emb_wt10",K = 65)

Single-embryo timing

To regenerate the single-embryo timing data underlying Figure 1, please run

# load metacell package
library("metacell")
# initializing the metacell scrna database
scdb_init("scrna_db")

source("scripts/single_embryo_timing.r")
gen_fig_1_plots()
embryo_ranks = gen_single_embryo_timing()

# subfunctions calculating intrinsic_rank and reference_rank of each embryo
# are contained in gen_single_embryo_timing()

The output data frame embryo_ranks was added to the single-cell metadata information of the metacell matrix object. All subsequent functions using single-embryo time information, are extracting it from the cell_metadata entry of the WT metacell single-cell matrix object sing_emb_wt10.

# load metacell package
library("metacell")
# initializing the metacell scrna database
scdb_init("scrna_db")

mat = scdb_mat("sing_emb_wt10")
md = mat@cell_metadata

Parameter stability analysis of network flow model

The parameter stability analysis of network flows underlying Figure S2A can be regenerated using

# regnerate data - this might take some time
source("scripts/parameter_stability_analysis.r")
gen_parameter_stability_analysis()

# replotting Figure S2A
source("scripts/generate_paper_figures/fig_s2.r")
fig_s2a()

Foxc12 chimera and tetraploid analysis

To generate specific plots of Figures 6, S6 and S7, please run the corresponding functions from fig_6.r, fig_s6.r or fig_s7.r. Users interested in recomputing parts of the Foxc12 chimera and tetraploid embryo analysis (not needed for regenerating the plots), should run the following functions:

library("metacell")
scdb_init("scrna_db/")

source("scripts/foxc12/generate_chimera_tetraploid_data_analysis.r")

# Chimera embryos injected with Foxc12 DKO cells
foxc_chimera_generate_time_and_cell_type_annotation()

# Chimera embryos injected with control cells
control_chimera_generate_time_and_cell_type_annotation()

# Tetraploid embryos injected with Foxc12 DKO cells
foxc_tetraploid_generate_time_and_cell_type_annotation()

# Tetraploid embryos injected with control cells
control_tetraploid_generate_time_and_cell_type_annotation()

This will transfer cell-type and time annotation from the wt atlas to chimera/tetraploid embryos. Output is saved in data/chimera_tetraploid_analysis/. Scripts involved in preprocessing plates from the chimera and tetraploid embryo analyis are saved in the scripts/foxc12/preprocessing/. This includes

• Gating of single cells using the FACS GFP channel
• Removing cells from extraembryonic ectoderm and parietal endoderm
• Merging each single-cell matrix with the wt single-cell matrix and creating a joint single-cell graph (metacell cgraph object). See summary_preprocessing.r and the corresponding scripts for more details.

Using Docker

To generate all the figures of the paper using the docker image, please run the following commands:

docker pull tanaylab/embflow:latest
mkdir figs
docker run -ti --user $(id -u):$(id -g) -v $(pwd)/figs:/embflow/figs tanaylab/embflow:latest

And then run within the R session:

source("scripts/initialize_scripts.r")
generate_all_figures()

The figures would be then generated in the mounted directory "figs".

Contact

For help, please contact [email protected]