Integrating cross-sample SVs with OctopusV

[sen1019san]

Hi OctopusV team,

Thanks for developing this great SV merging tool! I noticed OctopusV is designed to integrate SV results from different callers. I've generated SV calls for multiple samples using the TentacleSV pipeline, and now want to merge these sample-level SVs into a unified VCF file. Considering the same SV might have different IDs, breakpoints, or length variations across samples, could OctopusV handle this integration effectively? If yes, would you please suggest an example command for this use case?

Best regards,
Sen

[qingxiangguo]

Hi Sen,

Thanks for your question and for using OctopusV and TentacleSV!

Yes, OctopusV can effectively merge SVs across multiple samples, even if the IDs, breakpoints, or lengths vary slightly.

For your use case, you have two options:

• Use the --union strategy to combine SVs across all samples while automatically merging near-duplicate variants.
• Alternatively, use --min-support 2 to retain only SVs supported by at least two samples.

Example commands:

Union:

octopusv merge -i sample1.vcf sample2.vcf sample3.vcf -o merged_samples.vcf --union

Minimum support:

octopusv merge -i sample1.vcf sample2.vcf sample3.vcf -o merged_samples.vcf --min-support 2

Both methods will automatically merge close SVs based on default distance and size similarity thresholds.

Let us know if you have any more questions!

Best,

OctopusV team

[sen1019san]

Thank you for your prompt response, OctopusV team. I will try out the suggested solution.

[sen1019san]

Hello, could you explain why some SVs (Structural Variants) show a genotype of '0/0' after OctopusV integrates results from different SV callers?

[sen1019san]

Hello, I merged SV records from 12 samples using the command:

octopusv merge -i $sample_sv_records -o merged.vcf --union  

In the resulting merged.vcf, each SV entry only lists samples where the SV is present. For example, in the attached demo.txt, an SV appearing in 5 samples only displays those 5 samples' genotypes, rather than including genotypes for all 12 samples.

demo.txt

Furthermore, the SOURCES field (e.g., SOURCES=BTY.vcf,FZX.vcf,HML.vcf,KOM.vcf,SYH.vcf) does not match the genotype records in order. Specifically:

BTY.vcf should correspond to 0/1:11,16:27:352 but does not.
FZX.vcf should correspond to 1/1:0,25:25:351 but does not.
Expected behavior:

• Merged VCF should report genotypes for all input samples, using 0/0 for samples without the SV.
• SOURCES entries and genotype records should maintain a strict one-to-one correspondence.

A common scenario in merging sample VCF files involves parsing the "#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT SAMPLE" header to map sample names to genotypes accurately.
Resulting merged VCF header:
#CHROM POS ID ... FORMAT SMPLE001 SMPLE002 SAMPLE003

[qingxiangguo]

Hi, thanks for the detailed and thoughtful feedback!

You're absolutely right, currently, OctopuSV reports genotypes only for samples that support a given SV. Other samples are omitted rather than shown as 0/0. We recognize that for many downstream applications, explicitly reporting 0/0 is important. And we plan to add something like --include-absent-samples option to address this need.

Regarding the SOURCES field: it's currently designed to summarize which input VCFs contributed support for the SV, and is not in one-to-one correspondence with the genotype sample order. We’ll clarify this in the documentation, and are also considering a more explicit key-value format for better interpretability.

Thanks again, it really helps improve OctopuSV!

[sen1019san]

Thank you! Looking forward to OctopusV updates to support merging SV information across samples.