Author: Syed Nurul Hasan, Ph.D.
Affiliation: NYU Grossman School of Medicine (Ramasamy Lab)
Contact: nayanchoton1789@gmail.com | ORCID: 0000-0002-4564-872X
This repository demonstrates a complete end-to-end pipeline for multimodal single-cell analysis integrating scRNA-seq and scATAC-seq data from the same cells using 10x Genomics Multiome technology. The pipeline applies Weighted Nearest Neighbor (WNN) integration to jointly model gene expression and chromatin accessibility, enabling higher-resolution immune cell type identification than either modality alone.
The analytical framework here directly mirrors approaches used in the characterization of immune microenvironment dynamics in disease contexts — including leukemia, cardiovascular disease, and autoimmunity — where resolving transitional cell states at chromatin resolution is critical for identifying transcription factor (TF) networks driving immune cell reprogramming.
Single-cell ATAC-seq (scATAC-seq) captures chromatin accessibility genome-wide, providing a window into the regulatory landscape of individual cells. When integrated with scRNA-seq via WNN:
- Cells where chromatin is more informative receive higher ATAC weight
- Cells where transcription is more informative receive higher RNA weight
- The joint embedding resolves transitional immune states invisible to either modality alone
- TF motif deviation scores (chromVAR) identify which regulatory programs are active per cell
This is particularly powerful for questions like:
What transcription factor networks drive non-classical monocyte reprogramming in disease?
How does chromatin accessibility change in immune cells across disease progression?
Which regulatory programs are pharmacologically accessible?
10x Multiome Raw Data (PBMC)
│
▼
1. Quality Control & Filtering
├── RNA: nFeature, nCount, % mitochondrial
└── ATAC: TSS enrichment, nucleosome signal, peak counts
│
▼
2. RNA Processing (Seurat)
├── Normalization (SCTransform)
├── PCA dimensionality reduction
└── Initial RNA-based clustering
│
▼
3. ATAC Processing (Signac + ArchR)
├── Peak calling (MACS2)
├── LSI dimensionality reduction
├── chromVAR TF motif deviation scores
└── Gene activity score matrix
│
▼
4. WNN Integration (Seurat)
├── Weighted Nearest Neighbor graph
├── Joint UMAP embedding
└── WNN-based clustering
│
▼
5. Biological Interpretation
├── Immune cell type annotation
├── Monocyte subpopulation resolution
├── TF motif enrichment per cluster
├── Peak-to-gene linkage analysis
└── Cell-Cell Communication (CellChat)
│
▼
6. Visualization & Export
├── UMAP plots (RNA, ATAC, WNN)
├── TF deviation heatmaps
├── Coverage plots at key loci
└── Publication-ready figures
| Feature | Tool | Purpose |
|---|---|---|
| scRNA-seq processing | Seurat v5 | Normalization, PCA, clustering |
| scATAC-seq processing | Signac | Peak calling, LSI, chromatin QC |
| TF motif analysis | chromVAR | Per-cell TF deviation scores |
| WNN integration | Seurat v5 | Joint RNA+ATAC embedding |
| Gene activity | Signac | Imputed gene expression from ATAC |
| Peak-to-gene links | Signac | Regulatory element mapping |
| Cell communication | CellChat | Ligand-receptor network analysis |
| Visualization | ggplot2, Nebulosa | Publication-ready figures |
This pipeline uses the publicly available 10x Genomics PBMC Multiome dataset:
- 10,412 PBMCs from a healthy human donor
- Simultaneous scRNA-seq + scATAC-seq from same cells
- Download: 10x Genomics Dataset Portal
Why PBMC Multiome for immune biology:
PBMCs contain the same immune cell populations relevant to disease contexts — classical monocytes (CD14+), non-classical monocytes (CD16+), T cells, NK cells, B cells, and DCs. This dataset provides a clean baseline to validate multimodal integration approaches before applying them to disease specimens.
scATAC-WNN-multimodal/
├── README.md # This file
├── R/
│ ├── 01_quality_control.R # QC filtering for RNA and ATAC modalities
│ ├── 02_rna_processing.R # SCTransform, PCA, RNA clustering
│ ├── 03_atac_processing.R # Peak calling, LSI, chromVAR
│ ├── 04_wnn_integration.R # WNN graph, joint UMAP, WNN clustering
│ ├── 05_cell_annotation.R # Immune cell type annotation
│ ├── 06_tf_motif_analysis.R # TF deviation scores, motif enrichment
│ ├── 07_peak_gene_links.R # Peak-to-gene linkage analysis
│ └── 08_cellchat_communication.R # Ligand-receptor network analysis
├── docs/
│ ├── pipeline_guide.md # Step-by-step execution guide
│ └── biological_interpretation.md # Interpreting TF motif results
├── figures/ # Output figures directory
└── data/ # Data directory (see download instructions)
# Core packages
install.packages("Seurat") # v5.0+
install.packages("Signac") # v1.12+
BiocManager::install("chromVAR")
BiocManager::install("motifmatchr")
BiocManager::install("BSgenome.Hsapiens.UCSC.hg38")
BiocManager::install("JASPAR2020")
install.packages("CellChat")
# Visualization
install.packages("ggplot2")
install.packages("patchwork")
install.packages("viridis")
BiocManager::install("Nebulosa")
# Peak calling (system)
# conda install -c bioconda macs2# Clone repository
# git clone https://github.com/syed1789/scATAC-WNN-multimodal
# Run pipeline sequentially
source("R/01_quality_control.R")
source("R/02_rna_processing.R")
source("R/03_atac_processing.R")
source("R/04_wnn_integration.R")
source("R/05_cell_annotation.R")
source("R/06_tf_motif_analysis.R")
source("R/07_peak_gene_links.R")
source("R/08_cellchat_communication.R")While this pipeline is demonstrated on healthy PBMCs, it was developed to address specific questions in disease immunology:
Leukemia immune microenvironment:
Non-classical monocytes (CD14dim CD16+) expand at B-ALL diagnosis and relapse and predict inferior patient survival (Witkowski et al., Cancer Cell 2020). The chromatin-level determinants of this expansion — which TF networks drive NCM emergence and sustain their pro-leukemic phenotype — remain unmapped. This pipeline provides the analytical framework to address that question using paired scRNA-seq + scATAC-seq on immune fractions from matched diagnosis/remission/relapse specimens.
Cardiovascular immunology:
Myeloid cell state transitions in ischemic and diabetic hearts drive immune-metabolic pathway dysregulation. WNN integration of paired RNA and ATAC data resolves transitional myeloid states invisible to transcriptomics alone, enabling identification of TF networks driving pathological immune remodeling.
Autoimmunity:
Regulatory T cell (Treg) transcriptional programs — including Bcl11b-dependent chromatin accessibility — are disrupted in autoimmune disease. scATAC-seq of Treg populations enables chromatin-level characterization of tolerance versus dysfunction, with direct implications for leukemia-associated Treg immunosuppression.
- Stuart T, et al. (2021). Multimodal single-cell chromatin analysis with Signac. Nature Methods, 18:1333–1341.
- Hao Y, et al. (2021). Integrated analysis of multimodal single-cell data. Cell, 184:3573–3587.
- Schep AN, et al. (2017). chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nature Methods, 14:975–978.
- Witkowski MT, et al. (2020). Extensive remodeling of the immune microenvironment in B cell acute lymphoblastic leukemia. Cancer Cell, 37:867–882.
- Narang S, et al. (2024). Clonal evolution of the 3D chromatin landscape in patients with relapsed pediatric B-cell acute lymphoblastic leukemia. Nature Communications, 15:7425.
This pipeline is part of a broader single-cell genomics portfolio developed during postdoctoral research at NYU Grossman School of Medicine:
- scRNA-seq analysis: Seurat-based immune cell profiling in cardiac disease
- Spatial transcriptomics: VisiumHD analysis of tissue immune architecture
- Multi-omics integration: WNN pipeline for cardiac immune-metabolic interactions
- Cell-cell communication: CellChat network modeling of myeloid-stromal interactions
For questions or collaboration inquiries: nayanchoton1789@gmail.com