Multi-SC

Description

Our pipeline leverages a multimodal constraint autoencoder (scHCAE) to integrate the multiomics data during the clustering process and a matrix factorization-based model (scMF) to predict target genes regulated by a TF.

Architectural of scHCAE

Architectural of scMF

Usage

Prerequisites

• Python 3.8 or higher
• Required packages: torch, sklearn, scipy, scanpy, h5py, numpys, pandas

Setup

git clone https://github.com/xianglin226/Multi-SC.git

Running scHCAE

To run the scHCAE for clustering with a specified number of clusters, use the following command:

python -u run_scMultiCluster3.py \
 --n_clusters 6 \
 --data_file GSE178707_neatseq_lane1.h5

Running scMF

To predict target genes regulated by TFs using scMF, run:

python -u run_scMF.py \
 --data_file processedinput_scMF_lane1.h5

Example Data

The example data can be access here.

Clustering Data

GSE178707_neatseq_lane1.h5
GSE178707_neatseq_lane2.h5
GSM5123951_TEAseq_well1.h5

Structure

• X1: Gene expression data (RNA)
• X2: Protein expression data (ADT)
• X3: Chromatin accessibility data (ATAC)
• X4: Chromatin accessibility data (ATAC) mapped to gene features
• Genes: Gene features (rows of X1)
• ADT: Surface protein features (rows of X2)
• Peaks: Peak features (rows of X3)
• GeneFromPeaks: Gene features (rows of X4)
• Barcode: Cell barcodes

MF Data

processedinput_scMF_lane1.h5
processedinput_scMF_lane1.h5

• B: ADT-to-cell matrix
• W: Gene-to-ADT matrix
• X: Cell-to-gene matrix