SCRATCH-CNV Subworkflows

Introduction

This repository contains subworkflows for performing copy number variation (CNV) analysis on single-cell RNA-seq (scRNA-seq) data. The CNV subworkflows support multiple tools including inferCNV, SCEVAN, and CopyKAT to assess tumor heterogeneity and identify chromosomal aberrations.

Disclaimer: Subworkflows are high-level wrappers around chained Nextflow modules. They should be used as part of a pipeline and can be extended or reused across different scRNA-seq workflows.

Prerequisites

Ensure the following tools are installed:

• Nextflow (v21.04.0 or higher)
• Java (v8 or higher)
• Singularity or Docker for container execution
• Git

Installation

Clone the repository:

git clone https://github.com/WangLab-ComputationalBiology/SCRATCH-CNV.git
cd SCRATCH-CNV

Subworkflows

main.nf

This is the main entry script that orchestrates CNV analysis using inferCNV, SCEVAN, and CopyKAT subworkflows.

1. inferCNV

Performs CNV inference using gene expression intensities compared between reference (normal) and observation (tumor) cells.

Usage

nextflow run main.nf -profile singularity --input_seurat_object <path/to/seurat_object.RDS> --input_reference_table <path/to/reference_table.csv>

Parameters

• --input_seurat_object: Seurat object with UMAP and count layers
• --input_reference_table: CSV with barcode and reference label columns
• --project_name: Output project name (optional)
• --skip_infercnv: Skip running inferCNV (default: false)

2. SCEVAN

Performs CNV detection using Bayesian inference across multiple tumor samples.

Parameters (passed via ext.args)

• project_name: Name for output and figures
• input_model: Organism (e.g., human)
• n_threads: Number of threads
• n_memory: Memory in GB
• workdir: Working directory for outputs
• auto_save: Save intermediate objects (true/false)

3. CopyKAT

Optional module for an alternative CNV inference strategy.

Example

nextflow run main.nf -profile singularity \
  --input_seurat_object project_cluster_object.RDS \
  --input_reference_table assets/OV_reference_table.csv \
  --project_name OV_CNV \
  -resume

Annotated Object Requirement

To ensure successful CNV analysis, your input Seurat object must include one of the following annotation columns in meta.data and contain the minimum required cell types:

Annotated Metadata Requirements

Annotation Column	Required Cell Types	Role
azimuth_labels	B cell, T cell, Fibroblast, Epithelial	Reference + Observation
sctype	B_Plasma_Cells, T_Cells, Fibroblast, Epithelial	Reference + Observation
cell_label	B cell, T cell, Fibroblast, Epithelial	Reference + Observation

Note: The presence of these cell types is critical to define both reference (normal) and observation (tumor) populations.

Configuration

Default parameters and paths can be set in nextflow.config. Use institutional profiles for HPC environments.

Output

• ./<project_name>/data/infercnv: CNV matrices and plots from inferCNV
• ./<project_name>/data/scevan: CNV profiles and oncoheatmaps from SCEVAN
• ./<project_name>/report: Consolidated report

Notes on Troubleshooting

• For inferCNV HMM mode on large matrices, increase cutoff (e.g., 0.25) or use HMM_type = "i3" to reduce model complexity.
• Inspect logs via .nextflow.log and .command.out in work directories for failed tasks.
• Avoid missing parameters in ext.args block for modules like SCEVAN to prevent pipeline crashes.

Contributing

Open issues or submit PRs for bugs, enhancements, or suggestions.

License

This project is licensed under the GNU General Public License v3.0.

Contact

For help and questions:

• [email protected]
• [email protected]