Universal VCF Mutation Analysis Dashboard (UMAD)

Overview

The Universal VCF Mutation Analysis Dashboard (UMAD) is a bioinformatics pipeline designed to process and analyze large collections of Variant Call Format (VCF) files from cancer genome datasets.

The goal of UMAD is to provide a reproducible workflow for mutation characterization, comparison across samples, and biological interpretation of mutation patterns, in both OncoGAN synthetic and TCGA real data.

The pipeline currently processes simulated cancer datasets (Breast Adenocarcinoma and Prostate Adenocarcinoma)

These analyses support the development of an interactive dashboard for exploring mutation patterns across cancer samples.

Key Objectives

• Evaluate the biological realism of OncoGAN-generated genomes
• Compare mutation patterns between OncoGAN synthetic and real cancer datasets
• Validate biological plausibility using COSMIC mutational signatures
• Quantify differences in mutation burden and variability
• Provide an interactive dashboard for exploration and interpretation

Data Sources

• Synthetic Data: OncoGAN-generated VCF files
• Real Data: TCGA (Breast and Prostate cancer datasets)
• Reference: COSMIC Single Base Substitution (SBS) signature

1. Mutation Burden Analysis

• Computes total mutations per sample
• Compares distributions between synthetic and real datasets
• Statistical testing using Mann–Whitney U test
• Effect size (r) used to quantify magnitude of differences

2. Mutation Spectrum

• Classifies SNVs into six base substitution types
• Generates normalized mutation spectra
• Evaluates dominant mutation patterns (e.g., C>T transitions)

3. COSMIC Signature Analysis

• Extracts mutational signatures from combined spectra
• Compares against COSMIC reference signatures
• Generates heatmaps and exposure plots
• Assesses biological fidelity of synthetic data

Project Structure

BINF7700_Capstone_CCO/
├── dashboard.py
├── README.md
├── requirements.txt
│
├── src/                  # Core analysis logic
├── scripts/              # Helper shell scripts
│
├── data/                 # Input data
├── results/              # Processed outputs
├── figures/              # Final visualizations
│
├── logs/
├── docs/
└── cosmic_fit_output/

Methods

1. Mutation Burden

• Defined as total number of mutations per sample
• Log transformation applied to handle skewed distributions
• Compared between synthetic and real data

2. Statistical Analysis

• Mann-Whitney U test used to compare mutation distributions
• Rank-biserial correlation used to measure effect size

3. Mutation Spectrum

• Computed frequency of base substitutions (e.g., C>T, C>A)
• Used to identify dominant mutation patterns

4. Mutational Signatures

• COSMIC SBS signatures extracted and normalized
• Key signatures identified (SBS1, SBS5, SBS40, etc.)

5. Similarity Analysis

• Cosine similarity used to compare mutation profiles to COSMIC references

Requirements

• bcftools (v1.21+)
• Python 3.x
• Linux environment (HPC recommended)

Python Libraries

• pandas
• numpy
• scipy
• plotly
• streamlit

Key Findings

• OncoGAN data reproduces COSMIC mutational signatures, supporting biological plausibility
• Mutation spectra show consistent dominance of known patterns (e.g., C>T transitions)
• OncoGAN samples exhibit:
  • Higher mutation burden
  • Reduced variability
• Real datasets show:
  • Greater heterogeneity
  • Broader mutation distributions

These results highlight both the strengths and limitations of synthetic cancer genomes.

Dashboard

The Streamlit dashboard provides:

• Mutation burden visualization
• Mutation spectrum analysis
• COSMIC signature heatmaps
• Cosine similarity comparison
• Summary and interpretation

How to Run

git clone https://github.com/Naz-95-code/BINF7700_Capstone_CCO.git
cd BINF7700_Capstone_CCO
pip install -r requirements.txt
streamlit run dashboard.py

Author

Chinazo Christella Orji M.Sc. Bioinformatics Northeastern University, Toronto.