SRescue: Short- and Long-Read SV Integration Pipeline

A Nextflow pipeline for integrating short-read (SR) and long-read (LR) structural variant (SV) calls, rescuing SR-only SVs using LR data, and generating a polished, merged SV callset.

This pipeline is designed for tumor–normal WGS analyses and focuses on improving SV sensitivity by combining complementary sequencing technologies.

---

Workflow Overview

1. Merge LR and SR SVs

   • Merge LR and SR VCFs using SURVIVOR

2. Polish merged SVs

   • Normalize and convert merged VCFs to BEDPE/VCF formats

3. Classify SVs

   • Identify:
     • SR-only SVs
     • LR-only SVs
     • Shared SVs

4. Force calling SR-only SVs

   • Generate VCF for force calling
   • Re-call SR-only SVs in normal and tumor BAMs using cuteSV

5. Somatic filtering

   • Compare tumor vs normal calls
   • Filter and retain high-confidence somatic SVs
   • Collect supporting read evidence

6. Final integration

   • Merge rescued SR SVs with LR SVs
   • Final polishing, sorting, compression, and indexing

---

Inputs

• Long-read SV VCF
• Short-read SV VCF
• Tumor BAM (+ BAI)
• Normal BAM (+ BAI)
• Reference genome (FASTA)

---

Outputs

• final.sr2lr.polished.vcf.gz
• final.sr2lr.polished.vcf.gz.tbi

---

Dependencies

• Nextflow
• SURVIVOR
• cuteSV
• samtools
• bgzip / tabix
• Perl (custom polishing and filtering scripts)

---

Usage

nextflow run main.nf \
--lrvcf_fn longread.vcf \
--srvcf_fn shortread.vcf \
--norm_fn normal.bam \
--tum_fn tumor.bam \
--reference reference.fa