scMGCA 1.0.5

Topological Identification and Interpretation for High-throughput Single-cell Gene Regulation Elucidation across Multiple Platforms using scMGCA

scMGCA

scMGCA is a Python package containing tools for clustering single-cell data based on a graph-embedding autoencoder that simultaneously learns cell–cell topology representation and cluster assignments.

• Overview
• System Requirements
• Installation Guide
• Demo
• Data Availability
• License

Overview

Single-cell RNA sequencing (scRNA-seq) provides high-throughput gene expression information to explore cellular heterogeneity at the individual cell level. A major challenge in characterizing high-throughput gene expression data arises from the curse of dimensionality, and the prevalence of dropout events. To address these concerns, we developed a single-cell clustering method (scMGCA) based on a graph-embedding autoencoder that simultaneously learns cell–cell topology representation and cluster assignments. In scMGCA, we propose a graph convolutional autoencoder to preserve the topological information of cells from the embedded space in multinomial distribution, and employs the positive pointwise mutual information (PPMI) matrix for cell graph augmentation. Experiments show that scMGCA is accurate and effective for cell segregation and superior to other state-of-the-art models across multiple platforms, and is also able to correct for batch effects of data from different scRNA-seq protocols. In addition, we perform genomic interpretation on the key compressed transcriptomic space of the graph-embedding autoencoder to demonstrate the underlying gene regulation mechanism. In a pancreatic ductal adenocarcinoma (PDAC) dataset, with 57,530 individual pancreatic cells from primary PDAC tumors and control pancreases, scMGCA successfully provided annotations on the specific cell types and revealed differential gene expression levels across multiple tumor-associated and cell signalling pathways in PDAC progression through single-cell trajectory and gene set enrichment analysis.

System Requirements

Hardware requirements

scMGCA package requires only a standard computer with enough RAM to support the in-memory operations.

Software requirements

OS Requirements

This package is supported for Linux. The package has been tested on the following systems:

• Linux: Ubuntu 18.04

Python Dependencies

scMGCA mainly depends on the Python scientific stack.

numpy
scipy
tensorflow
scikit-learn
pandas
sklearn

requirements in https://github.com/Philyzh8/scMGCA

Installation Guide:

Install from PyPi

$ conda create -n scMGCA_env python=3.6.8
$ conda activate scMGCA_env
$ pip install -r requirements.txt
$ pip install scMGCA

Demo

Usage

You can run the scMGCA from the command line:

$ from scMGCA.run import train
$ train(data, highly_genes=500, pretrain_epochs=1000, maxiter=300)

Arguments

Parameter	Introduction
data	A h5 file. Contains a matrix of scRNA-seq expression values,true labels, and other information. By default, genes are assumed to be represented by columns and samples are assumed to be represented by rows.
highly genes	Number of genes selected
pretrain epochs	Number of pretrain epochs
maxiter	Number of training epochs

Example

$ from scMGCA.run import train
$ data = './dataset/Quake_10x_Bladder/data.h5'
$ train(data, highly_genes=500, pretrain_epochs=1000, maxiter=300)

The result will give NMI and ARI

Data Availability

The real data sets we used can be download in data.

License

This project is covered under the MIT License.