Single-Cell RNA Analysis Pipeline

This pipeline provides tools for preprocessing and quality control analysis of single-cell RNA sequencing data. It performs initial quality filtering, generates diagnostic plots, and annotates genes of interest.

Features

• Quality control filtering based on:
  • Mitochondrial gene percentage
  • Number of genes per cell
  • Number of cells expressing each gene
  • Outlier detection using Median Absolute Deviations (MAD)
• Doublet detection and filtering
• Highly variable gene selection
• Gene expression visualization
• Custom gene set analysis
• Comprehensive quality metrics and plots

Installation

1. Clone the repository:

git clone https://github.com/yourusername/pipe_scanalysis.git
cd pipe_scanalysis

2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Unix/macOS
# or
.\venv\Scripts\activate  # On Windows

3. Install dependencies:

pip install -r requirements.txt

Usage

The pipeline can be run using the run_qc_process.py script. Here's the basic command structure:

python run_qc_process.py --input_counts_file <input.h5> --output_counts_file <output.h5ad> [options]

Required Parameters

• --input_counts_file: Path to the input AnnData file containing raw counts
• --output_counts_file: Path where the processed AnnData file will be saved

Optional Parameters

• --output_dir: Output directory for plots and statistics (default: current directory)
• --geneset_ids_file: File containing a list of gene IDs to analyze
• --mad_outlier_threshold: MAD threshold for outlier filtering (default: 5)
• --mito_percentage_threshold: Maximum allowed mitochondrial gene percentage (default: 8)
• --genes_exp_in_min_cells_threshold: Minimum number of cells expressing each gene (default: 0)
• --cells_with_min_genes_threshold: Minimum number of genes per cell (default: 0)
• --n_top_expr_genes: Number of top expressed genes to plot (default: 50)
• --n_highly_variable_genes: Number of highly variable genes to select (default: 5000)
• --dispersion_flavor: Method for highly variable gene selection (default: 'seurat_v3_paper')
  • Options: 'seurat_v3_paper', 'seurat_v3', 'seurat', 'cell_ranger'
• --gene_ids_of_interest_file: File containing gene IDs for additional quality metrics
• --normalisation_method: Normalization method to use (default: 'shifted')
  • Options: 'shifted', 'pearson'
• --drop_doublets: Flag to filter out predicted doublets

Example Usage

python run_qc_process.py \
    --input_counts_file unfiltered_gex_counts.h5 \
    --output_counts_file filtered_counts.h5ad \
    --output_dir results \
    --mito_percentage_threshold 10 \
    --cells_with_min_genes_threshold 200 \
    --n_highly_variable_genes 3000 \
    --drop_doublets

Output

The pipeline generates:

1. A filtered and annotated AnnData file with processed data
2. Quality control plots and statistics in the specified output directory

Dependencies

The pipeline requires Python 3.x and the following main packages:

• scanpy
• anndata
• numpy
• pandas
• matplotlib
• scikit-learn
• umap-learn

For a complete list of dependencies and their versions, see requirements.txt.

License

See MIT License

Contributing

[Add contribution guidelines here]