siLOH - 🛢️🔎 SNP Inferred Loss of Heterozygosity Analysis Pipeline

siLOH is a Docker-based pipeline for analysing Loss of Heterozygosity (LOH) from NGS data. The pipeline integrates samtools, VarScan2, and a custom Python analysis script to identify potential regions of Loss of Heterozygosity (LOH).

The siLOH pipeline uses a targeted approach to identify candidate LOH regions by analysing the allele frequencies of predefined common SNPs (curated SNPs from dbSNP with Minor Allele Frequency (MAF) ≥ 30%) across the capture regions contained within an input BAM file.

Important Notes on Usage ❗

This tool is designed for research use and preliminary screening purposes only. Results should be considered indicative rather than diagnostic:

• All LOH findings should be confirmed using validated diagnostic methods
• Negative results do not rule out the presence of LOH
• The tool is optimised for specificity over sensitivity to minimise false positives
• Results should be interpreted in conjunction with other clinical and laboratory findings
• This tool should not be used as the sole basis for diagnostic or treatment decisions

For clinical applications, follow-up testing using validated diagnostic methods (such as SNP arrays, microsatellite analysis, or targeted sequencing) is required to confirm any findings.

Pipeline Workflow

Loading

flowchart TD
    %% External inputs
    A[Input BAM File]
    C[Reference Genome]
    
    subgraph "Docker Container"
        direction TB
        
        subgraph "Built-in Resources"
            B[MAF30 SNPs List]
            cent[Centromeres.json]
            bed[Gene BED File]
        end

        subgraph "Step 1: Variant Calling"
            D[Samtools mpileup]
            E[VarScan2 pileup2cns]
            F[Raw Variants at SNP Sites]
        end

        subgraph "Step 2: LOH Analysis"
            G[Analyze Variant Frequencies]
            H{Is Position Homozygous?<br/>≤35% or ≥65%}
            I[Build Region Streak]
            J{Region Criteria Met?}
            K[Split at Centromeres]
        end

        subgraph "Step 3: Filtering & Output"
            L{Final Region Filters}
            M[Gene Annotation]
        end
    end
    
    %% Output file
    N[LOH Report CSV]

    %% Connections outside Docker
    A --> D
    C --> D
    M --> N
    
    %% Connections inside Docker
    B --> D
    D --> E
    E --> F
    F --> G
    G --> H
    H -->|Yes| I
    H -->|No| G
    I --> J
    J -->|No| G
    J -->|"Yes<br/>≥5 homozygous sites<br/>≥1Mb size"| K
    K --> L
    cent --> K
    L -->|"Pass:<br/>≥40 homozygous<br/>>90% confidence<br/>non-X chromosome"| M
    bed --> M

    %% Styling
    classDef input fill:#e3f2fd,stroke:#1565c0
    classDef docker fill:#f5f5f5,stroke:#424242,stroke-width:2px
    classDef resource fill:#e8eaf6,stroke:#3949ab
    classDef process fill:#f5f5f5,stroke:#424242
    classDef decision fill:#fff3e0,stroke:#ef6c00
    classDef output fill:#e8f5e9,stroke:#2e7d32

    class A,C input
    class B,cent,bed resource
    class D,E,F,G,I,K,M process
    class H,J,L decision
    class N output

Pipeline Steps

1. Pileup Generation: Uses samtools mpileup to generate pileup data at specified positions
2. Variant Calling: VarScan2 analyzes the pileup to identify variants
3. LOH Analysis: Custom Python script processes variants to identify LOH regions
4. Output Generation: Results are saved in CSV format

Prerequisites

• Docker installed and running
• Input BAM files
• Reference genome (ucsc_hg19.fa)
• BED file containing regions of interest
• Sufficient disk space for analysis

Installation

1. Clone the repository:

git clone https://github.com/g-pyxl/siLOH.git
cd siLOH

2. Build the Docker image:

docker build -t siloh .

Required Files

The pipeline expects the following files:

• Reference Genome: ref/ucsc_hg19.fa (GRCh37 example)
• Sample BAM Files: Contig naming should follow "chr1" format
• BED File: beds/R210.bed - The pipeline comes provided with an R210.bed designed for Lynch Syndrome calling. Tool can be ran with or without BED file calling.
• maf30.txt: Over 1.8m SNPs with high MAF, curated from dbSNP
• centromeres.json: Centromere positions relative to GRCh37 (provided in repo)

Directory Structure

siLOH/
├── Dockerfile
├── requirements.txt
├── run_analysis.sh
├── loh.py
├── maf30.txt
├── centromeres.json
├── ref/
│   └── ucsc_hg19.fa
├── samples/
│   └── your_sample.bam
├── results/
└── beds/
    └── R210.bed

Usage

1. Run the analysis:

docker run -v /path/to/ref:/app/ref \
           -v /path/to/samples:/app/samples \
           -v /path/to/results:/app/results \
           -v /path/to/beds:/app/beds \
           siloh your_sample_name

Replace /path/to/ with your actual paths and your_sample_name with your BAM filename (without the .bam extension).

Output

The pipeline generates the following files in the results directory:

• {sample}.pileup: Raw pileup data
• {sample}.cns: VarScan2 consensus output
• {sample}.loh.csv: Final LOH analysis results

The LOH CSV file contains:

• Chromosome
• Start position
• End position
• Affected genes (if BED file provided)

Debugging

To enter the container without running the analysis:

docker run -it --entrypoint=/bin/bash siloh

Resource Requirements

• Memory: Depends on input BAM file size
• Disk Space: ~3x the size of input BAM file
• CPU: Single-threaded processing

Known Limitations

• Currently supports hg19 reference genome
• Single-sample processing
• Requires sorted, indexed BAM files - .bai should be within same directory as BAM
• Employs a high-specificity approach that trades sensitivity for reliability:
  • Designed for expedited screening rather than comprehensive LOH detection
  • Will only detect more obvious/extensive LOH regions to minimize false positives
  • May miss subtle or complex LOH events that would be detected by traditional methods
  • Less sensitive than traditional non-NGS methods like SNP arrays or microsatellite analysis

Citation

Citation

Required Citations

This tool incorporates several open-source tools and methodological approaches that should be cited:

Tools

• Li H, Handsaker B, Wysoker A, et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009;25(16):2078-2079. doi:10.1093/bioinformatics/btp352

• Koboldt DC, Zhang Q, Larson DE, et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res. 2012;22(3):568-576. doi:10.1101/gr.129684.111

Methodological Foundations

• Tuna M, Knuutila S, Mills GB. Uniparental disomy in cancer. Trends Mol Med. 2009;15(3):120-128. doi:10.1016/j.molmed.2009.01.005

• Ryland GL, Doyle MA, Goode D, et al. Loss of heterozygosity: what is it good for? BMC Med Genomics. 2015;8:45. doi:10.1186/s12920-015-0123-z

Related Methods

For comparison with traditional approaches:

• Takahashi S, Fukuda M, Mitani Y, et al. Microsatellite instability and LOH studies for assessment of mismatch repair deficiency in colorectal cancer. Methods Mol Biol. 2021;2265:147-162. doi:10.1007/978-1-0716-1209-5_11

• González S, Jover L, Mila M, et al. Cost-effectiveness Analysis Comparing Different Techniques for MSI and LOH Studies in Lynch Syndrome Diagnosis. Appl Immunohistochem Mol Morphol. 2017;25(10):720-727. doi:10.1097/PAI.0000000000000370

License

Contact

Contributing