EchoSV is a versatile tool for comparing and merging structural variant (SV) call sets generated using different reference genomes. It studies how SVs "echo" across these references through a hybrid workflow that combines liftover and graph-based matching.

Given two or more SV call sets from the same sample—each aligned to a different reference—EchoSV can perform two primary operations:

• Compare: Generates a detailed comparison identifying overlapping variants and those exclusive to a specific reference, e.g., calls across GRCh38, CHM13, and a donor-specific assembly (DSA).
• Merge: Consolidates multiple SV call sets into a single, unified output, e.g., merging two DSA haplotype-based call sets into one consolidated file.

• Requirements
• Installation
• Usage
• License
• Contact

EchoSV depends on the following Python packages:

• pysam: Read and write BAM/CRAM files and VCF records for variant processing.
• intervaltree: Efficiently store and query genomic intervals to detect overlapping SVs.
• Biopython (Bio): Parse and manipulate sequence data during liftover steps.
• scipy: Perform statistical analysis and numerical computations on variant metrics.
• networkx: Construct and traverse graphs that model structural variant matches.
• pandas: Tabular data manipulation for comparison outputs.
• numpy: Numerical computations on variant metrics.
• rich: Formatted terminal output and progress display.

Option 1: From GitHub (recommended)

git clone git@github.com:parklab/EchoSV.git
cd EchoSV
pip install -r requirements.txt
pip install .

Option 2: Via PyPI

pip install echosv

The EchoSV workflow consists of four main steps: chain, merge (optional), genotype, and match. Below are detailed instructions and examples using the test data (can be downloaded from Zenodo).

Download the EchoSV test data echosv_test_data.tar.gz from Zenodo and decompress it:

tar -xzvf echosv_test_data.tar.gz

The chain command generates a liftover chain file that maps coordinates from ref2 (the source assembly) to ref1 (the target reference). Before running chain, align ref2 against ref1 using minimap2's asm-to-asm mode and index the output:

minimap2 -a -x asm5 --cs ref1.fa ref2.fa \
    | samtools view -hSb - \
    | samtools sort -O BAM -o ref2_to_ref1.bam
samtools index ref2_to_ref1.bam

Then generate the chain file. EchoSV looks for a pre-built index automatically to parse the contig lengths; if none exist, the FASTA is parsed directly (slower for large assemblies). You can generate an index with samtools faidx ref2.fa or samtools dict ref2.fa > ref2.fa.dict.

echosv chain \
    -b test_data/input_data/chm13_to_grch38.bam \
    -f test_data/input_data/chm13.fa \
    -o test_data/chm13_to_grch38.chain.gz

# Merge multiple VCFs from the same reference into a single call set
echosv merge \
    -i grch38_colo829_caller1.vcf.gz grch38_colo829_caller2.vcf.gz [...] \
    -o grch38_colo829_svs.vcf.gz \
    --merge --new

# Extract high-confidence SVs (≥4 supporting callers, ≥2 platforms)
echosv merge \
    -i grch38_colo829_svs.vcf.gz \
    -o grch38_colo829_svs_highconf.vcf.gz \
    --extract 

The genotype command collects supporting reads for each SV from BAM files and annotates the VCF with allele-frequency and read-name fields used by the graph-based matching in Step 4.

echosv genotype --longread \
    -i test_data/input_data/grch38_colo829_somatic_svs.vcf.gz \
    -b test_data/input_data/chm13_to_grch38.bam \
    -o test_data/grch38_colo829_genotyped.vcf.gz

The match command compares SV call sets across different reference genomes using a two-step hybrid approach: liftover-based coordinate matching followed by graph-based matching on shared supporting reads (echo score).

# Compare SV call sets and report concordant / reference-exclusive variants
echosv match -i test_data/test_colo829_config.json

# Compare SV call sets between DSA haplotypes and also produce a merged DSA-based VCF
echosv match -i dsa_merge_colo829_config.json --merge 

The input is a JSON config file specifying reference labels, genotyped VCFs, chain files, and the output path. See test_data/test_colo829_config.json below for a working example.

{
    "refs":   { "1": "grch38", "2": "chm13", "3": "dsa" },
    "vcfs":   { "1": "./test_data/grch38_colo829_genotyped.vcf.gz",
                "2": "./test_data/chm13_colo829_genotyped.vcf.gz",
                "3": "./test_data/dsa_colo829_genotyped.vcf.gz" },
    "chains": { "2_to_1": "./test_data/chm13_to_grch38.chain.gz",
                "3_to_1": "./test_data/colo829bl_hap*_grch38.chain.gz" },
    "output": "./test_data/colo829_svs_comparison.txt"
}

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

Feel free to open an issue on GitHub or contact Yuwei Zhang (yuwei_zhang@hms.harvard.edu) if you have any questions about EchoSV.