New pipeline: nf-core/<vasta>

Pipeline title/name

vasta

Keywords

structural variants, whole genome sequecing, genomics, machine learning

What is it about?

Structural Variant Calling and Trust Assessment using VarTrustML (vasta) is a confidence-driven pipeline for structural variant analysis that combines multi-caller detection with machine learning-based assessment of individual structural variant (SV) reliability.

At its core, the pipeline incorporates VarTrustML, a newly developed Python library that estimates the reliability of individual SV calls by integrating caller support with genomic context, sequencing coverage, and alignment-derived signatures.

This approach moves beyond conventional caller-consensus strategies, enabling the generation of high-confidence SV callsets while retaining potentially true variants supported by a single caller.

SV and CNV detection, quality control, callset normalization, annotation, and ML-based confidence scoring are integrated into a standardized and reproducible nf-core workflow.

Please provide a schematic diagram of the proposed pipeline

What would a minimal first release of this pipeline include?

The minimal first release would include all components required to generate high-confidence structural variant callsets from short-read whole-genome sequencing data.

This includes quality control (mosdepth and MultiQC), multi-caller SV detection (Manta, Delly, and Smoove), filtering of variants smaller than 50 bp, callset merging, and VCF normalization to standardize variant representation, INFO annotations, and FORMAT fields across callers.

The workflow would further perform extraction of genomic and alignment-derived features, followed by confidence scoring using VarTrustML.

The primary output would be a standardized SV VCF annotated with confidence scores and accompanied by comprehensive quality reports.

Additional modules, including CNV detection, functional annotation, and downstream variant prioritization, would be incorporated in future releases.

I confirm my proposed pipeline will follow nf-core guidelines. Most importantly, my pipeline will:

• be built with Nextflow.
• pass nf-core lint tests and use standardized parameters.
• be community-owned and developed within the nf-core organization.
• open source under the MIT license with proper credits and acknowledgments.
• have a descriptive, all lowercase, and without punctuation name.
• use the nf-core pipeline template and predominantly use official nf-core modules.
• focus on a specific data/analysis type with appropriate scope.
• have properly maintained documentation.
• be bundled using versioned Docker/Singularity containers.

Why do we need a new pipeline?

Structural variant detection from short-read whole-genome sequencing remains challenging, as no single caller can accurately identify all variant types and size ranges with high sensitivity and precision. Consequently, combining multiple callers through union or consensus strategies has become common practice. However, these approaches inevitably involve a trade-off between sensitivity and precision and do not directly estimate the reliability of individual variant calls.

vasta addresses this limitation by introducing a confidence driven framework that combines multi-caller SV detection with machine learning-based reliability assessment. Instead of relying solely on caller agreement, the pipeline extracts genomic context, sequencing coverage, and alignment-derived features to estimate the confidence of each structural variant using VarTrustML. This enables the generation of standardized, confidence-scored SV callsets while retaining potentially true variants that would otherwise be discarded by consensus-based approaches.

To our knowledge, no existing nf-core pipeline currently provides an integrated workflow for SV discovery coupled with machine learning–based confidence scoring. VASTA therefore provide a reproducible, modular, and scalable pipeline for high-confidence SV analysis from short-read whole-genome sequencing data.

Who would be interested?

Researchers working on short reads data

What has been done so far

A working pipeline has already been developed and is routinely used for the internal analysis of short-read whole-genome sequencing data.

The current implementation includes multi-caller SV detection (Manta, Delly, and Smoove), quality control, callset filtering, merging and normalization, genomic and alignment-derived feature extraction, and confidence scoring through the VarTrustML Python library.

VarTrustML has been benchmarked using publicly available reference datasets (e.g. GIAB) as well as internally curated datasets.

A manuscript describing the methodology and benchmarking results is currently in preparation. The main objective of this proposal is to refactor the existing implementation into a modular, fully reproducible, and community-maintained nf-core workflow

URL to existing work (if applicable)

https://github.com/emanuela04/vasta

Are there any similar existing nf-core pipelines?

No response

Start by reviewing the existing implementation at github.com/emanuela04/vasta and the proposal’s minimal first-release scope. Confirm how the workflow should use Manta, Delly, Smoove, VarTrustML, and the stated QC and normalization steps; done means a modular nf-core workflow that passes nf-core lint and produces the standardized, confidence-scored SV VCF and quality reports.