Lobular Coordinate Inference from scRNA-seq

Overview

This repository is dedicated to the inference of lobular coordinates from single-cell RNA sequencing (scRNA-seq) data. Our focus is on implementing and comparing three distinct computational approaches to address this challenge. This repository contains the code used in the study: A sexually dimorphic hepatic cycle of periportal VLDL generation and subsequent pericentral VLDLR-mediated re-uptake.

Code for the article

1. MLE.ipynb it's the notebook that assigns the lobular coordinate to cells based on their transcriptome. Such coordinate is present in several figures of the paper.
2. gene_selection.ipynb This notebook expands the list of zonated genes, leveraging the coordinates found with the smaller gene list.

Additional code

1. other_methods/numpyro_SVI.ipynb: Utilizing the NumPyro library, which is built on JAX, SVI offers a scalable Bayesian inference method, approximating posterior distributions of latent variables.
2. other_methods/numpyro_MCMC.ipynb: This approach, also using NumPyro, employs the No-U-Turn Sampler (NUTS), to sample the posterior distributions of all latent variables.
3. MLE_droin.ipynb: Same method applyed to another dataset from Droin et al. Space-time logic of liver gene expression at sub-lobular scale.

Configuration

The config.py file is central to the setup, containing key hyperparameters such as:

• List of genes used for analysis
• Number of iterations
• Batch size

Getting Started

In order to run the code the following packages are needed:

• python
• numpy
• matplotlib
• pip
• scikit-learn
• ipykernel
• pandas
• seaborn
• anndata
• scanpy
• python-igraph
• leidenalg
• scipy

For PyTorch approaches:

• torch
• torchvision
• torchaudio

For JAX-based approaches (SVI, MCMC):

• jax
• jaxlib
• numpyro

Contact

Contact [email protected]