Repository navigation
Description
Question
What is the supported way to run sentieon-cli dnascope-hybrid on a human sample with both diploid and haploid regions (for example, an XY sample)? Is the absence of --haploid_bed from the Hybrid CLI intentional, a documentation gap, or a missing feature/bug?
We are prototyping mixed-ploidy support in a downstream fork, but want to avoid introducing scientifically incorrect behavior.
Current documented/source behavior
As of the current upstream main (1bf377d3ce79fc4d8c2dc221e1f696441e38349d):
- The Hybrid documentation exposes only
-b DIPLOID_BED. dnascope_hybrid.pydefines only-b/--bedand uses that shared BED for Hybrid small-variant calling and its CNV path.- The DNAscope LongRead documentation separately documents
-b DIPLOID_BED --haploid_bed HAPLOID_BEDfor mixed-ploidy samples. - The CNVscope application note says male samples should use separate
CNVscopeandCNVModelApplypasses over disjoint diploid and haploid BEDs, followed by VCF combination. - Sentieon 202503.04 notes improved haploid CNVscope calling.
Relevant documentation:
- https://support.sentieon.com/docs/sentieon_cli/#dnascope-hybrid
- https://support.sentieon.com/docs/appnotes/cnvscope/#using-cnvscope-with-male-samples
- https://support.sentieon.com/docs/appendix/releasenotes/
Downstream behavior we are considering
For an XY sample, our candidate fork would accept:
sentieon-cli dnascope-hybrid \
... \
-b hg38_diploid_male_no_alt.bed \
--haploid_bed hg38_haploid_male_no_alt.bed \
sample.vcf.gz
The candidate implementation would:
- Require
-bwhenever--haploid_bedis supplied. - Run one diploid CNVscope -> CNVModelApply pass restricted to the diploid BED and one haploid pass restricted to the haploid BED, using the same
cnv.modelmember from the composite Hybrid bundle. - Combine those disjoint CNV VCFs in reference order.
- Run the Hybrid LongReadSV component over the sorted/merged union of the diploid and haploid BEDs.
- Refuse to run Hybrid small variants/gVCF with
--haploid_beduntil their supported haploid semantics and model contract are confirmed.
For XX samples, it would retain one diploid pass using the female diploid BED.
Questions for Sentieon
- Does current Hybrid
-bintentionally restrict all Hybrid outputs to only those intervals, meaning non-PAR X/Y are omitted when the official male diploid BED is supplied? - Is there an existing supported or planned
--haploid_bedcontract fordnascope-hybrid? - For Hybrid CNV, is the two-pass scheme above correct with the
cnv.modelfromHybridIlluminaONT2.0.bundle? ShouldCNVModelApplyreceive an interval too, or only the already interval-restricted CNVscope VCF? - For Hybrid LongReadSV, is using the union of diploid and haploid BEDs correct? Does LongReadSV have any ploidy-sensitive genotype behavior that requires the regions to remain separate?
- What is the supported approach for Hybrid SNV/indel and gVCF calling in haploid non-PAR X/Y? Can the Hybrid model be used in separate haploid processing, or is another model/pipeline required?
- Is Sentieon 202503.04 required for the supported male Hybrid behavior?
In short: would adding --haploid_bed with the behavior above close a gap, or would any part of it induce incorrect calls? We would be happy to shape an upstream PR around the supported contract once clarified.
Activity
iamh2o commentedon Aug 27, 2026
A concrete review branch is now available at https://github.com/iamh2o/sentieon-cli/tree/codex/hybrid-ploidy-components-1.7.3i (commit
4218c04). It implements the proposed two-pass CNVscope path and LongReadSV region union, but deliberately rejects Hybrid small-variant/gVCF execution when--haploid_bedis present. We have not tagged or used it for production calls; confirmation of the supported Hybrid haploid small-variant/model contract remains the release gate.DonFreed commentedon Aug 29, 2026
Sentieon version 202503.04 (which was recently released) updated
CNVscopeto explicitly support CNV calling across both diploid and haploid chromosomes. Here is the new command line usage:Additionally, version 202503.04 introduced support for a new hybrid pangenome pipeline. Development and model training for this new pipeline is currently in-progress (#33). For Human samples, the new pipeline can be used with the HPRC pangenome. Compared to the
dnascope-hybridpipeline, the pangenome pipeline has significant runtime and accuracy improvements.The next version of the
sentieon-cliwill update CNVscope calling (across all pipelines) to use the new--sexand--pararguments introduced in Sentieon v202503.04 to perform correct handling of male/female samples.LongReadSV and the DNAscope hybrid model currently only support diploid variant calling. If you are using separate diploid and haploid bed files, I would pass the union of the diploid and haploid bed files to these variant callers.
DonFreed commentedon Sep 27, 2026
Closing this for now. Feel free to re-open if you have additional questions.