- Jupyter Notebook 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| benchmark/dynamo | ||
| data | ||
| notebook | ||
| tools | ||
| .gitignore | ||
| README.md | ||
Figures of ddHodge
This repository provides the Jupyter notebooks that reproduce the results in our ddHodge paper.
The Julia package of ddHodge is available at ddHodge.jl.
Preface
Most of the code written here was devoted to making plots.
In addition to reproducing our results, these helper tools can also be used for simulation and data analysis using Julia.
- [tools/Dynamical2D.jl] : A Julia module for setting up 2D dynamical systems, such as calculating analytical div, grad and curl in curved space.
- [tools/H5ADHelper.jl] : A Julia module to add a new column to the
obsDataFrame of the loaded anndata file. - [tools/dynamo_nopca.py] : A Python script to run Dynamo in the original dimension (i.e., without dimension reduction). See notebook/dynamo_fig1_ess.ipynb and benchmark/dynamo for generating the input H5AD files.
Please refer to the examples below for specific usage:
- [tools/Dynamical2D.jl] : notebook/surfsim_fig.ipynb
- [tools/H5ADHelper.jl] : notebook/vasa_fig.ipynb
The basic usage of H5ADHelper.jl is also explained in ddHodge.jl.
Figures
These figures were produced using ddHodge (version 0.7.1).
Figure 1
Showing an example of artificial dynamics in 3D and performing velocity embedding:
Figure 2 and the related ones
Validation of ddHodge performance using an artificial 2D gradient system and toggle-switch model:
Figure 3 and the related ones
Validation of ddHodge in higher dimensional dynamics, such as artificial dynamics on a sphere, Lorenz system, feedback loop system and 100 genes RNA velocity simulation (reproduced from Gorin et. al.):
Figure 4-5 and the related ones
Using ddHodge to analyse scRNA-seq data:
The RNA velocities were calculated as follows:
The RNA velocity was calculated using scvelo (version 0.3.2).
Data availability
For Figures 1-3, the notebooks are self-contained. No additional files are required.
For Figures 4-5, some of the public data are needed to reproduce our results. The small size (< 20MB) of the pre-processed data has been uploaded to the data directory.
The processed data including ddHodge results of VASA-seq and regenerating muscle tissue datasets are available at Figshare.
The original datasets not included in this repository are availabel at:
- Spliced/unspliced read count of VASA-seq: GSE176588
- List of histone genes (HistoneGenes.tsv): VASAseq_2022
- FUCCI dataset (RNAData/a.loom) in: input.zip (Note: the file was renamed to
fucci_RNAData.loomin our analysis) - Regenerating muscle tissue dataset (raw sequence reads and metadata): GSE143437