Sovereign 8s.is mirror of standardgalactic/ddHodge_figures
  • Jupyter Notebook 100%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-02-25 14:53:19 +09:00
benchmark/dynamo Updates for revision 2025-09-24 21:10:58 +09:00
data Updates for revision 2025-09-24 21:10:58 +09:00
notebook Updates for revision 2025-09-24 21:10:58 +09:00
tools Updates for revision 2025-09-24 21:10:58 +09:00
.gitignore first commit 2025-03-05 10:06:55 +09:00
README.md Update paper reference in README 2026-02-25 14:53:19 +09:00

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 obs DataFrame 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:

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:

Validation of ddHodge performance using an artificial 2D gradient system and toggle-switch model:

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.):

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.loom in our analysis)
  • Regenerating muscle tissue dataset (raw sequence reads and metadata): GSE143437