jtb_2023_code

This repository contains figure generation code for the manuscript "Simultaneous estimation of gene regulatory network structure and RNA kinetics from single cell gene expression".

Model code is in the supirfactor-dynamical package. Velocity and time inference code is in the inferelator-velocity package. Sequencing data is deposited in NCBI GEO under accession GSE242556.

Overview

This project is the analysis and plotting layer for the manuscript. It packages processed data, computes shared derived quantities such as projections and decay summaries, loads model outputs from external packages, and renders the main and supplemental figures.

The repository is organized around a small Python package, jtb_2023_code, plus checked-in figure assets and helper scripts.

Repository layout

• /home/runner/work/jtb_2023_code/jtb_2023_code/jtb_2023_code: main Python package
  • generate_figures.py: top-level figure generation pipeline
  • preprocess.py: builds the processed shared dataset used by the figures
  • package_data.py: exports a packaged AnnData object with counts, embeddings, velocities, and decay values
  • figure_1.py through figure_6.py: main figure builders
  • figure_*_supplemental.py: supplemental figure builders
  • figure_constants.py: file paths, plotting constants, selected genes, and other central configuration
  • utils/: shared data loading, plotting, projections, decay, pseudotime, and model helper code
  • pseudotime/: wrappers for different pseudotime methods
  • old_data/: support for older published datasets used in later figures
• /home/runner/work/jtb_2023_code/jtb_2023_code/Data: reference tables shipped with the repository
• /home/runner/work/jtb_2023_code/jtb_2023_code/Figures: rendered main and supplemental figure outputs
• /home/runner/work/jtb_2023_code/jtb_2023_code/Schematic: static schematic artwork used inside the figures
• /home/runner/work/jtb_2023_code/jtb_2023_code/scripts: standalone runners for pseudotime workflows

Main workflow

At a high level, the figure pipeline does the following:

1. Loads single-cell expression data into a shared FigureSingleCellData object.
2. Computes or loads PCA, UMAP, pseudotime, decay, and related derived data.
3. Loads model outputs and predictions from external modeling packages.
4. Builds the manuscript figures and writes them to the figure output directory.

The main orchestrator is jtb_2023_code/generate_figures.py, which preprocesses the dataset and then calls each figure module in sequence.

Installation

The package metadata is defined in /home/runner/work/jtb_2023_code/jtb_2023_code/setup.py.

Primary Python dependencies include:

• numpy
• scipy
• scanpy
• pandas
• joblib
• anndata
• matplotlib
• pydeseq2
• supirfactor-dynamical
• inferelator-velocity

Install the repository in an environment that already has the scientific Python stack available:

pip install -e /home/runner/work/jtb_2023_code/jtb_2023_code

Data and path configuration

The code expects several data, figure, scratch, schematic, and model paths. These defaults are defined in jtb_2023_code/figure_constants.py and jtb_2023_code/utils/figure_filenames.py.

Some defaults point to author-specific local or scratch locations, so running the pipeline in a new environment usually requires overriding paths on the command line. The path helper supports:

• -d for the data directory
• -f for the figure directory
• -scratch for scratch files
• -s for the schematic directory
• -m for the model directory

Usage

Examples of the main entry points:

Preprocess the shared figure dataset

python /home/runner/work/jtb_2023_code/jtb_2023_code/jtb_2023_code/preprocess.py \
  -d /path/to/data \
  -scratch /path/to/scratch

Generate all figures

python /home/runner/work/jtb_2023_code/jtb_2023_code/jtb_2023_code/generate_figures.py \
  -d /path/to/data \
  -f /path/to/figures \
  -scratch /path/to/scratch \
  -s /path/to/schematics \
  -m /path/to/models

Package data for downstream use

python /home/runner/work/jtb_2023_code/jtb_2023_code/jtb_2023_code/package_data.py \
  -d /path/to/data \
  -scratch /path/to/scratch \
  --output_file /path/to/output/data.h5ad

Notes

• This repository focuses on manuscript analysis and figure production rather than the core dynamical model implementation.
• The checked-in Figures/ directory provides examples of the expected outputs.
• The scripts/ directory contains method-specific pseudotime runners and may require additional environment setup depending on the tool being used.

License

This repository is distributed under the MIT License. See /home/runner/work/jtb_2023_code/jtb_2023_code/LICENSE.