scMultiome: Single-Cell Multimodal Integration Pipeline for scRNA-seq and scATAC-seq with Gene Regulatory Network Inference

Max | GitHub: junior1p | https://github.com/junior1p/scMultiome

Abstract

scMultiome is a complete end-to-end Python pipeline for integrating paired single-cell RNA sequencing (scRNA-seq) and assay for transposase-accessible chromatin sequencing (scATAC-seq) data from multiome platforms (10x Multiome, SHARE-seq, SNARE-seq). The pipeline combines scGLUE (graph-linked unified embedding) and MOFA+ (multi-omics factor analysis) for multimodal dimensionality reduction, marker-based cell type annotation validated across both modalities, and cis-regulatory gene regulatory network (GRN) inference via GLUE embedding cosine similarity.

1. Introduction

Single-cell multi-omics enables simultaneous profiling of gene expression and chromatin accessibility from the same individual cell, eliminating computational cell-matching problems inherent in unpaired datasets. 10x Genomics Multiome, SHARE-seq, and SNARE-seq generate such paired data at scale.

Two state-of-the-art integration approaches address this differently: scGLUE leverages genomic coordinate proximity (peaks near genes) as a biological knowledge graph prior, while MOFA+ learns latent factors without external prior information.

2. Pipeline Architecture

2.1 Data Loading

Accepts: (A) automatic PBMC 10k demo download, (B) user-provided 10x .h5 or .h5mu, or (C) separate RNA and ATAC .h5ad files. Uses MuData framework throughout.

2.2 Quality Control

• RNA: gene count (200–7500), mitochondrial fraction <20%
• ATAC: peak count (1000–30000), total counts (2000–100000)
• Intersection: only cells present in both modalities proceed

2.3 Preprocessing

• RNA: normalize → log1p → HVG selection (3000) → scale → PCA (30)
• ATAC: TF-IDF → LSI (truncated SVD on 30k HVPs) → PCA-equivalent LSI

2.4 Multimodal Integration

Two complementary methods:

• scGLUE: genomic proximity knowledge graph (1 Mb window) → shared latent embedding
• MOFA+: Bayesian factor model for shared + modality-specific factors

2.5 Cell Type Annotation

Canonical PBMC markers scored per cell; highest-scoring label assigned. ARI checks cross-modal consistency.

2.6 GRN Inference

GLUE embeddings place genes and peaks in shared vector space → cosine similarity >0.5 identifies cis-regulatory peak–gene pairs → optional TF motif scanning via JASPAR.

3. Installation

pip install muon scanpy scglue anndata mofapy2 leidenalg     python-igraph matplotlib seaborn pandas numpy scipy     --break-system-packages -q

4. Usage

from multiome import run_multiome_skill

mdata, metrics, grn = run_multiome_skill(
    input_path=None,  # Downloads PBMC 10k automatically
    out_dir="results",
    run_scglue=True,
    run_mofa=True,
    run_grn=True,
    max_epochs=200
)

5. Output Files

6. Dependencies

muon≥0.1.6, scanpy≥1.9.6, scglue≥0.3.3, anndata≥0.10.0, mofapy2≥0.7.1, leidenalg≥0.10.1, python-igraph≥0.11.0, matplotlib≥3.7, seaborn≥0.12, pandas≥1.5, numpy≥1.24, scipy≥1.10, scikit-learn≥1.3, requests≥2.28

Python 3.9+. GPU (CUDA) optional but recommended for scGLUE.

References

1. Bredikhin, D. et al. (2022). MUON: multimodal omics analysis framework. Genome Biology.
2. Cao, Z.-J. & Gao, G. (2022). Multi-omics single-cell data integration and regulatory inference with graph-linked embedding. Nature Biotechnology.
3. Hao, Y. et al. (2021). Integrated analysis of multimodal single-cell data. Cell.