Biosynthetic Gene Cluster (BGC) Tools Comparison

📋 Table of Contents

• Overview
• Key Features
• Installation
• Project Structure
• Tools Analyzed
• Results
  • Streptomyces coelicolor
  • Streptomyces ameniacus
  • Streptomyces avidinii
• Contributing

🔍 Overview

This project provides a detailed comparative analysis of different tools used for detecting and classifying Biosynthetic Gene Clusters (BGCs) in bacterial genomes. The analysis focuses on three important Streptomyces species and evaluates the performance of leading BGC detection tools: antiSMASH, GECCO, and DeepBGC.

✨ Key Features

• Comprehensive comparison of three major BGC detection tools
• Analysis of genomic data from three Streptomyces species
• Visual representation of tool overlaps using Venn diagrams
• Detailed statistical analysis of tool performance
• Jupyter notebook implementation for reproducible research

🚀 Installation

1. Clone this repository:

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

2. Install the BGC detection tools:

# make an Environment
conda create -n bgc_detection python=3.8
conda activate bgc_detection

# antiSMASH
pip install antismash

# GECCO
pip install gecco-tool

# DeepBGC
pip install deepbgc

3. Run the Jupyter notebook:

jupyter notebook main.ipynb

📁 Project Structure

The project follows this directory structure for tool outputs:

BGC_Prediction/
├── S.coelicolor/
│   ├── antismash/     # antiSMASH output files
│   ├── gecco/         # GECCO output files (including clusters.tsv)
│   └── deepbgc/       # DeepBGC output files
├── S.ameniacus/
│   ├── antismash/
│   ├── gecco/
│   └── deepbgc/
└── S.avidinii/
    ├── antismash/
    ├── gecco/
    └── deepbgc/

Each tool's output is organized in separate directories for each genome, making it easy to access and compare results.

🛠️ Tools Analyzed

antiSMASH

• Industry standard for BGC detection
• Provides detailed annotations and classifications
• Comprehensive analysis of secondary metabolite biosynthesis pathways

Running antiSMASH: The analysis was performed using the antiSMASH 6.0 web server:

1. Visit https://antismash.secondarymetabolites.org
2. Upload genome files in GenBank/FASTA format
3. Use default settings for bacterial genome analysis
4. Download the GenBank result files

GECCO

• Rule-based approach for gene cluster detection
• Focuses on genomic context and gene organization
• Efficient processing of large genomic datasets

Running GECCO:

# For each genome file (in FASTA format):
gecco run \
    --genome input_genome.fasta \
    --output gecco_output \
    --proteins proteins.faa \  # Optional: provide protein sequences
    --threads 4 \
    --force    # Overwrite existing output directory

GECCO generates three output files in the specified output directory:

• clusters.tsv: Contains the predicted BGC locations and types (used for our analysis)
• features.tsv: Detailed information about genomic features
• proteins.faa: Protein sequences in FASTA format

DeepBGC

• Machine learning-based BGC classification
• Utilizes deep learning models for prediction
• Capable of detecting novel BGC types

Running DeepBGC:

# Download the pre-trained model (if not already present)
deepbgc download

# Detect BGCs in genome sequences
deepbgc pipeline \
    --input input_genome.fasta \
    --output deepbgc_output \
    --label genome_name \
    --detector hybrid \
    --classifier product-class

Important Notes

• All tools were run with their default parameters to ensure fair comparison
• Input genomes were provided in FASTA format for GECCO and DeepBGC
• antiSMASH results were downloaded in GenBank format
• All analyses were performed on complete genome sequences

📊 Results

Data Format

The analysis results are stored in two main formats:

1. TSV Summary Files

Located in the BGC_predictions folder, these files contain the comparative analysis results:

Column	Description
Sequence ID	Identifier for the genomic sequence
Start	Start position of the predicted BGC
End	End position of the predicted BGC
BGC Type	Predicted type of the biosynthetic gene cluster
antiSMASH	Binary indicator (1/0) for antiSMASH detection
GECCO	Binary indicator (1/0) for GECCO detection
DeepBGC	Binary indicator (1/0) for DeepBGC detection

2. GenBank Files

Located in the output_gbk folder, these files contain detailed genomic annotations:

File	Description	Size
S.coelicolor_A3_modified.gbk	Annotated genome of S. coelicolor	19MB
Streptomyces_ameniacus_modified.gbk	Annotated genome of S. ameniacus	20MB
Streptomyces_avidinii_modified.gbk	Annotated genome of S. avidinii	17MB

The GenBank files contain:

• Complete genome sequences
• Gene annotations
• Predicted BGC locations and types
• Detailed feature annotations from each tool
• Protein and coding sequences
• Tool-specific metadata and predictions

These files can be viewed with genome browsers or bioinformatics tools like Artemis, UGENE, or SnapGene.

Region Intersection Logic

BGC regions are considered to intersect when they:

1. Share the same sequence ID
2. Have overlapping genomic coordinates

The intersection is determined by the following Python function:

def regions_intersect(region1, region2):
    sequence_id1, start1, end1 = region1
    sequence_id2, start2, end2 = region2
    return (str(sequence_id1) == str(sequence_id2)) and (max(start1, start2) <= min(end1, end2))

This means two regions intersect if:

• They are on the same sequence (sequence_id1 == sequence_id2)
• The start of one region is before the end of the other region (max(start1, start2) <= min(end1, end2))

Comparative Analysis

Streptomyces coelicolor

Key findings:

• Total BGCs detected: 162
• 21 BGCs detected by all three tools (13%)
• 115 unique regions identified by DeepBGC (70%)
• 9 unique regions detected by GECCO (5%)
• 2 unique regions found by antiSMASH (1%)
• Most common BGC types: Polyketide, RiPP, NRP

Streptomyces ameniacus

Key findings:

• Total BGCs detected: 118
• 20 BGCs detected by all three tools (17%)
• 81 unique regions identified by DeepBGC (68%)
• 5 unique regions detected by antiSMASH (4%)
• 2 unique regions found by GECCO (2%)
• Most common BGC types: Polyketide, NRP, Terpene

Streptomyces avidinii

Key findings:

• Total BGCs detected: 106
• 23 BGCs detected by all three tools (21%)
• 67 unique regions identified by DeepBGC (62%)
• 8 unique regions detected by GECCO (7%)
• 1 unique region found by antiSMASH (1%)
• Most common BGC types: Polyketide, NRP, Terpene

Key Observations

1. Tool Agreement: Approximately 15-20% of BGCs are detected by all three tools across species
2. DeepBGC Sensitivity: DeepBGC consistently identifies the most unique regions (60-70%)
3. BGC Types:
   • Polyketide clusters are the most frequently detected
   • NRP (Non-Ribosomal Peptide) clusters are common across all species
   • Terpene clusters show high conservation
4. Unknown Regions: DeepBGC identifies many regions as "Unknown" type, suggesting potential novel BGCs

🤝 Contributing

Contributions are welcome :) Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request