scPN 0.1.5

A package for single cell pseudotime

scPN: Simultaneous Inference of Pseudotime and Gene Interaction Networks

Overview

scPN is a framework that simultaneously infers pseudotime and gene-gene interaction networks from scRNA-seq data. The framework integrates clustering, piecewise linear modeling, and an iterative EM-style algorithm to recover both temporal dynamics and regulatory relationships among genes.

Figure 1. The Framework of scPN

Figure 1. The framework of scPN, which can simultaneously obtain the temporal dynamics and the gene-gene interaction matrix.

• (a) Raw dataset of gene expression. The single-cell gene expression matrix is typically of size ( T imes N ), where ( T ) represents the number of cells and ( N ) denotes the number of genes. This matrix is often sparse.
• (b) The preprocessing procedure of scPN. It includes normalization, gene selection, imputation, clustering, piecewise linear network modeling, and initialization of the gene-gene interaction matrix using prior knowledge.
• (c–d) Constructing individual piecewise networks after clustering. scPN clusters cells using the Leiden algorithm and constructs distinct piecewise gene regulatory networks for each cluster, corresponding to different time intervals.
• (e) scPN algorithm. The iterative algorithm, similar to the Expectation-Maximization (EM) algorithm, alternates between inferring pseudotime via a TSP-based approach and estimating the interaction matrix using regression.
• (f) Output of scPN. Outputs include single-cell pseudotime, velocity fields, and a gene-gene interaction matrix. Further downstream analysis can be conducted on the learned regulatory networks.

🔧 Requirements

To use scPN, you need to install the following Python packages:

pip install scanpy scvelo numpy torch matplotlib

🚀 Usage

To run the demo, simply run the cells in Test&Contrast one by one.

Ensure your working directory contains the input data.

📫 Contact

For questions or suggestions, feel free to open an issue or contact the authors.