SwiftCNV

SwiftCNV is a fast and scalable Python implementation of the core InferCNV algorithm to infer copy number variations (CNVs) from single-cell RNA-seq data. It provides additional features and is designed for seamless interoperability with Anndata objects and Scanpy.

Documentation

For detailed information and example tutorials, please refer to our documentation.

Installation

SwiftCNV can be installed through pip:

pip install swiftcnv

Python dependencies are: numpy, pandas, scipy, scikit-learn, anndata, matplotlib

Extra dependencies for the tutorial: scanpy, ipykernel, leidenalg, requests

pip install swiftcnv[tutorial]

Usage

SwiftCNV can be run from the command line from a h5ad file, but can also be imported to your script for advanced usage and AnnData/Scanpy integration. SwiftCNV requires a portion of the cells to be defined as reference to calculate the CNV score, normally being cells that are not expected to be malignant (e.g., based on cell type). The program requires a GTF file containing gene annotations, which be obtained from the Gencode database.

Command Line Interface

Required arguments

Optional arguments

Reference cells can be specified using a TSV file with two columns, cell_name and reference, where the reference column contains TRUE or FALSE to indicate whether each cell is used as a reference. Alternatively, reference cells can also be specified by providing the column in adata.obs containing the cell type annotations (--reference-col) and which ones should be used as reference (--reference-value). Finally, the column identifying the samples must be specified.

Example

A typical call from a cells file would be:

swiftcnv \
    -i /path/to/adata.h5ad \
    -o /path/to/output \
    -c /path/to/cells.tsv \
    -a /path/to/gene_annotations.gtf.gz \
    -s sample \
    -p \
    --hmm \
    --exclude-immune

Where the required cells.tsv file (-c / --cells) would be:

Alternatively, one can use cell types from the input file to define the reference

swiftcnv \
    -i /path/to/adata.h5ad \
    -o /path/to/output \
    -a /path/to/gene_annotations.gtf.gz \
    --reference-col cell_type \
    --reference-vals T_cell Macrophage Fibroblast \
    -s sample \
    -p \
    --hmm

Outputs

Output files are placed under -o/--output (or output_dir if defined). SwiftCNV generates 3 main files:

• cnv_scores.npz: compressed matrix (cells x genes) containing the CNV values
• cell_order.tsv.gz: file containing cell barcodes, reference status and sample_id if provided
• gene_order.tsv.gz: file containing gene metadata

If -p/--plot was specified:

• cnv_scores.png: Heatmap plot of the whole CNV scores matrix
• cnv_scores_by_sample.pdf: Heatmap plots of each sample if provided

if --hmm was specified HMM segmentation outputs will go to a hmm/ directory:

• cnv_states.tsv.gz: DataFrame containing 3-states labels matrix by subcluster, cell or sample, depending on --hmm-by
• cnv_states.png: Heatmap plot with the found HMM states
• tumor_subclusters.tsv.gz: subcluster labels for the state HMM clustering if --hmm-by=subcluster (default)