Sample-scRNAseq-Pipeline 🧬

A comprehensive single-cell RNA sequencing (scRNA-seq) analysis pipeline demonstrating quality control, dimensionality reduction, clustering, cell type annotation, and differential expression analysis using modern Python bioinformatics tools.

📋 Overview

This repository contains a complete Jupyter Notebook workflow for analyzing 10X Genomics single-cell RNA sequencing data. The pipeline covers the full analysis workflow, from raw count matrices to annotated cell populations, including differential gene expression analysis.

Key Analysis Steps:

• Loading and preprocessing 10X Genomics data (barcodes, features, matrix)
• Quality control and filtering (mitochondrial content, gene counts)
• Data normalization and variable gene selection
• Dimensionality reduction with scVI (deep learning-based)
• UMAP visualization and Leiden clustering
• Automated cell type annotation using CellTypist
• Differential expression analysis across clusters
• Custom visualization and data export

✨ Features

• Modern scRNA-seq Stack: Leverages Scanpy, scVI-tools, and CellTypist
• Deep Learning Integration: Uses scVI for robust dimensionality reduction and batch correction
• Automated Cell Typing: CellTypist integration for rapid cell type annotation
• Transcription Factor Focus: Filters data to focus on transcription factor genes
• Quality Visualizations: Generates publication-ready plots (UMAP, violin plots, heatmaps)
• Reproducible Workflow: Complete end-to-end analysis in a single notebook
• Model Persistence: Saves trained scVI models for reproducibility

🚀 Installation

Prerequisites

• Python 3.9+ (tested on Python 3.10)
• Jupyter Notebook or JupyterLab
• ~2GB RAM minimum for the example dataset
• GPU recommended (but not required) for scVI training

Setup

1. Clone the repository:

git clone https://github.com/deep-kapadia-6/sample-scRNAseq.git
cd sample-scRNAseq

2. Create a conda environment (recommended)

conda create -n scrnaseq python=3.10
conda activate scrnaseq

3. Install required dependencies:

pip install -r requirements.txt

The following packages will be installed:

• anndata - Annotated data structures for single-cell data
• matplotlib - Data visualization
• mudata - Multi-modal data handling
• muon - Multi-omics analysis framework
• scanpy - Single-cell analysis in Python
• scvi - Deep generative models for single-cell omics
• numpy - Numerical computing
• pandas - Data manipulation

💻 Usage

Running the Analysis

1. Launch Jupyter Notebook:

jupyter notebook scRNAseq_code.ipynb

2. Prepare your data files: Place 10X Genomics output files in the appropriate directory:

• file_barcodes.tsv.gz
• file_features.tsv.gz
• file_matrix.mtx.gz

Update file paths in the notebook to match your data location

3. Execute cells sequentially to run the full analysis pipeline

Input Data Format

The pipeline expects standard 10X Genomics output:

• Barcodes: Cell barcodes (one per row)
• Features: Gene identifiers (one per row)
• Matrix: Sparse count matrix (Market Exchange Format)

Customization

Filter Thresholds (Cell 16):

sc.pp.filter_cells(rna_adata, min_genes=200)  # Minimum genes per cell
sc.pp.filter_genes(rna_adata, min_cells=3)    # Minimum cells per gene
rna_adata = rna_adata[rna_adata.obs.pct_counts_mt < 40, :]  # Mitochondrial content

scVI Model Parameters:

model = scvi.model.SCVI(adata)  # Default: n_latent=10, n_hidden=128
model.train()  # Default: 400 epochs

Clustering Resolution:

sc.tl.leiden(adata, resolution=0.5, key_added="leiden_scVI")  # Adjust resolution

📂 Project Structure

sample-scRNAseq/ ├── scRNAseq_code.ipynb # Main analysis notebook ├── requirements.txt # Python dependencies ├── README.md # This file ├── LICENSE # MIT License └── .gitignore # Git ignore file

🔬 Analysis Workflow

1. Data Loading & QC

• Load 10X data into AnnData object
• Calculate QC metrics (genes per cell, UMI counts, mitochondrial %)
• Visualize distributions before filtering

2. Preprocessing

• Filter low-quality cells and genes
• Normalize counts (CPM normalization)
• Log-transform data
• Filter for transcription factor genes (optional)

3. Dimensionality Reduction (scVI)

• Train variational autoencoder on count data
• Generate 10-dimensional latent representation
• Save trained model for reproducibility

4. Clustering & Visualization

• Compute k-nearest neighbors graph
• Leiden clustering algorithm
• UMAP for 2D visualization
• Force-directed graph layout (ForceAtlas2)

5. Cell Type Annotation

• Automated annotation using pre-trained CellTypist models
• Majority voting for cluster-level annotation
• Confidence scoring for predictions

6. Differential Expression

• Identify marker genes for each cluster
• t-test or logistic regression methods
• Rank genes by statistical significance

7. Visualization & Export Generate publication-quality plots

Export annotated AnnData object (.h5ad format)

📊 Example Output

The pipeline generates several key visualizations:

• QC Plots: Violin plots showing genes/cell, UMI counts, and mitochondrial percentage
• UMAP Plots: Colored by cluster ID, cell type annotation, or gene expression
• Dotplots: Cluster composition and annotation confidence
• Heatmaps: Top marker genes per cluster
• Pie Charts: Cell type proportions

🛠️ Technical Details

Computational Requirements

• Memory: ~2-4 GB for small datasets (<5k cells), scales with data size
• Runtime: ~5-15 minutes for full pipeline (depends on dataset size and CPU/GPU)
• GPU: Optional but recommended for faster scVI training

Key Technologies

• Scanpy: Industry-standard scRNA-seq analysis framework
• scVI-tools: State-of-the-art probabilistic models for single-cell genomics
• CellTypist: Automated cell type annotation using machine learning
• AnnData: Efficient storage format for annotated data matrices
• PyTorch: Deep learning backend for scVI

Data Specifications

• Cells: Designed for 1k-10k cells (scalable to larger datasets)
• Genes: Full transcriptome or subset (TF-focused in example)
• Format: Sparse matrix (memory-efficient for scRNA-seq data)

📝 Citation

If you use this pipeline, please cite the underlying tools:

• Scanpy: Wolf et al. (2018) Genome Biology
• scVI: Lopez et al. (2018) Nature Methods
• CellTypist: Domínguez Conde et al. (2022) Science
• AnnData: Virshup et al. (2021) bioRxiv

🤝 Contributing

Contributions are welcome! Areas for improvement:

• Add support for multi-sample integration
• Implement trajectory inference analysis
• Add RNA velocity analysis
• Create a command-line interface
• Add automated report generation
• Improve documentation with example datasets
• Please open an issue or submit a pull request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Analysis pipeline adapted from best practices in single-cell genomics
• Developed for research in leukemia immunobiology and stem cell biology
• Built using open-source bioinformatics tools from the Python ecosystem

📧 Contact

Created by @deep-kapadia-6

For questions about the analysis or to report issues, please open a GitHub issue.

Note: This pipeline is for research purposes only. Ensure you have appropriate permissions and ethical approvals before analyzing human genomic data.