Single-cell multiomics of human fetal hematopoiesis defines a developmental specific population and a fetal signature

N. B. Before running any scripts or notebooks, please edit the file project_path to contain your local path to this folder. For example, if your username is bioinf and this folder (directory) is called Sommarin_et_al and resides in your home directory, the project_path file should contain:

/home/bioinf/Sommarin_et_al

The scripts and notebooks read this file to set working directory, output directories etc, so please do this first. You can also use direnv, see the included .envrc file (which also reads the project_path file). This will set the environment variable PROJECT_PATH to be what you specified in the project_path file.

N. B. II The procedure to run this project is an amalgamation of approaches, due to such things as different code authors with different backgrounds, several revision rounds etc. To abstract, the pipeline for running this project is as follows:

1. Run the Seurat part
2. Run the 1st notebooks part
3. Run the DESeq2 part
4. Run the 2nd notebooks part

One level down abstraction-wise:

1. Activate the conda environment specified in envs/seurat.yaml and run the src/seurat-analysis/run_all.sh script (working directory/where to execute this script from: $PROJECT_PATH/src/seurat-analysis).
2. Activate the conda environment specified in envs/notebooks.yaml and run the associated notebooks with this command: jupyter nbconvert --inplace --execute --config config_for_running_notebooks_pre_DESeq2.py (working directory $PROJECT_PATH/notebooks)
3. Install snakemake (e.g. in a conda environment) and run the DESeq2 part with this command: snakemake -j all --use-conda (working directory $PROJECT_PATH).
4. Activate the conda environment specified in envs/notebooks.yaml and run the associated notebooks with this command: jupyter nbconvert --inplace --execute --config config_for_running_notebooks_post_DESeq2.py (working directory $PROJECT_PATH/notebooks)
5. Gather the produced figures used for the paper by running snakemake again, letting it find the scattered plots and copying them into the figures/figures folder. Use the same command as above: snakemake -j all --use-conda (working directory $PROJECT_PATH).

Directory structure

Guide for folder (directory) names:

• data
  • contains the data generated in and needed for this study.
  • external: external data used
  • processed: the output of this project (data and figures)
  • interim: "middle-steps" that are used to generate processed data but is in itself not part of the final results
  • raw: the raw data generated in this study
• envs
  • contains conda environment specifications for the three different parts of this project. Also used by snakemake for the DESeq2 part of this study.
  • install conda (or even better, mamba), and then each environment can be generated with the command conda env create -f <environment_file.yaml> (substitute conda with mamba if you installed mamba). (Don't forget to also activate each environment after creation before running any scripts or notebooks.)
    • (note I: if you already have an environment with the same name, you can e.g. change the name in the first line of corresponding .yaml file to avoid conflict errors.)
    • (note II: snakemake will create the environments it needs for itself from these files, but you will still need to create them for the seurat and notebooks parts of the study.)
• figures
  • directory where all the project's plots that are used as figures for the accompanying paper are gathered. This is done by snakemake. The naming convention for the figures files is: <figure index>_<original filename>. E.g. Figure1b_UMAP_FL.svg
• notebooks
  • jupyter notebooks (if you generate environments like specified above, jupyterlab is installed)
  • both .ipynb and .py representations are present, they are synced with jupytext. You can either run the .py files or the .ipynb files. For executing the .ipynb (notebook) files, including generating cell output, there are two files for that: config_for_running_notebooks_pre_DESeq2.py and config_for_running_notebooks_post_DESeq2.py. The files are not meant to be run themselves, see the first lines of comments in the files for a command to run (which uses the respective file as a config file). Also included is a command to trigger the syncing of .py and .ipynb files after changes made to any of the formats (syncing will nevertheless happen automatically if you use jupyter from the environment included in this project (envs/notebooks.yaml) to open an .ipynb file).
• src
  • scripts used in this study
  • seurat-analysis: scripts for the part using the R package Seurat (amongst others). There is a shell script for running all R scripts: run_all.sh. Activate the seurat environment and run this. You might have to allow it to run as well (e. g. chmod +x run_all.sh).
  • DESeq2: scripts for the part using the R package DESeq2 (amongst others). Use snakemake to run this part.

This project is runnable in its current form on Unix-based systems (Linux or macOS), i. e. not Windows. The paths are specified with a forward slash. If you are using Windows and want to reproduce these results, one suggestion would be to use WSL2 (Windows Subsystem for Linux).

Docker container

Another approach is to use a Docker container. We have prepared one that should enable all parts of this analysis to run out of the box. You can find it on dockerhub at https://hub.docker.com/r/razofz/fetal-liver. When you have docker installed you can pull (download) the container with this command:

docker pull razofz/fetal-liver:0.4

Depending on how you configured Docker you might have to preface the command with sudo. See e. g. here for some pointers on how to configure Docker for use without sudo.
When you have pulled the container you can start an instance of it with the following command. If you use this exact command make sure you have changed the project_path file to contain simply /fl. Note that this will reflect the path on the container and not on your computer. You also need to have navigated to the directory on your computer in which this project resides for this to work (due to the PWD command in the command). In other words, if you have downloaded this project as in the example above, you need to have navigated to /home/bioinf/Sommarin_et_al to run the following command:

docker run --interactive --tty --volume ${PWD}:/fl --rm --name Sommarin_et_al razofz/fetal-liver:0.4 /bin/bash

This will give you a bash session in the Docker container. The project files are in the /fl dir, and you can follow the instructions above on activating the conda environments (they have already been created in this container, see the included Dockerfile for details).

/Rasmus

P.S. Note that our adult bone marrow samples have the label yBM in many of the scripts. This is due to that data being from a study where there was even older bone marrow data, so in relation to that these were the younger bone marrow data. Also note that our CS22 data have label hpc at times. D.S.

---

(Project structure is a slimmed-down version of the Data Science Cookiecutter)