Single-cell transcriptomic profiling presents itself as a powerful tool to understand biological development.
In a 2024 publication, Qiu et al. leveraged the single-cell RNA-seq data of mouse embryo nuclei spanning late gastrulation to birth to infer a rooted tree of cell-type relationships. In this repo, we adapted the code from Qiu et al. to delineate the relationship between gut cells and hepatocytes. Our work shows that similar cell-type lineage could be inferred for any cell types or tissues of interest.
The R scripts require the following R packages:
- Matrix
- dplyr
- ggplot2
The Python script requires the following Python libraries:
- scanpy
- pandas
- numpy
- annoy
Run the scripts in the following order:
- extract_subset_cells.R
- subset_scRNAseq_wf.py
- visualization.R
"extract_subset_cells.R" extracts data of given cell types from the complete mtx files. It probably requires at least 100GB of RAM. "subset_scRNAseq_wf.py" reprocesses the extracted subset of data, applying dimensionality reduction using scanpy. "visualization.R" makes 2D UMAP plots for the selected cell types, with the option of highlighting the expression of a specific gene.
To run these scripts successfully, you need to download the two folders named "mtx" and "other" from the JAX data (see link in the Acknowledgement section). Make sure to change the paths to where you want to save and load the data, including the intermediate files (e.g., lines 7-8 of "extract_subset_cells.R").
Here is a 2D UMAP plot with the mutual nearest neighbors in gut cells and hepatocytes highlighted.
Yellow circles represent gut cells bearing the highest resemblance to hepatocytes, whereas purple circles
denote hepatocytes bearing the highest resemblance to gut cells. The separate "island" in the bottom left
corner mostly consists of hepatocytes at birth.
Here is a 2D UMAP plot with Gata6 expression levels highlighted. Gata6 is a master regulator
dictating hepatic development.
Junmin Wang