Feature: Evaluate GENOMICON-Seq for realistic PCR duplicate simulation

[berntpopp]

Summary

Evaluate GENOMICON-Seq as a potential simulation module to generate biologically realistic PCR duplicates with cycle-by-cycle amplification modeling—addressing the sampling duplicate limitation in Wessim (see #51).

Background

GENOMICON-Seq is a modern exome simulation tool that explicitly models both the capture step and PCR amplification in a unified pipeline:

Feature	Wessim (current)	GENOMICON-Seq
Capture bias	✅ Probe hybridization	✅ Probe capture efficiency
PCR duplicates	❌ Sampling-based	✅ Cycle-by-cycle amplification
Error model	External (GemSim/ART)	Integrated per-cycle errors
Duplicate families	Binary (2 copies)	Realistic clonal expansion

How GENOMICON-Seq works

1. Probe capture simulation: Models which fragments are "pulled down" based on hybridization efficiency
2. PCR amplification: Uses an inverted sigmoid function to model amplification efficiency per cycle
3. Error introduction: Adds noise during PCR cycles, creating realistic duplicate families with shared errors

This produces PCR duplicate families where one source molecule → N copies with correlated errors, matching wet-lab behavior.

Availability

• GitHub: https://github.com/Gabaldonlab/GENOMICON-Seq
• Docker: Available as container
• Publication: GENOMICON-Seq paper (if applicable)

Investigation Tasks

Phase 1: Feasibility assessment

• Clone and test GENOMICON-Seq standalone
• Verify Docker container works in our CI environment
• Assess input/output compatibility with MucOneUp pipeline
• Check license compatibility (MucOneUp is MIT)

Phase 2: Integration evaluation

• Map GENOMICON-Seq parameters to MucOneUp CLI options
• Determine if it can use custom FASTA (our corrected haplotypes)
• Evaluate probe file requirements (SureSelect, IDT, custom?)
• Benchmark runtime vs Wessim at 100x coverage

Phase 3: Duplicate quality comparison

# Compare duplicate characteristics
# GENOMICON-Seq output
picard MarkDuplicates I=genomicon_output.bam M=genomicon_metrics.txt

# Wessim output (from #51)
picard MarkDuplicates I=wessim_output.bam M=wessim_metrics.txt

# Compare family size distributions

Integration Options

Option A: Add as new simulator backend

muconeup run --simulator genomicon-seq --coverage 100x

• Pros: Preserves backward compatibility, user choice
• Cons: Maintenance burden of multiple backends

Option B: Replace Wessim for Illumina exome

• Pros: Simplified codebase, better defaults
• Cons: Breaking change, may affect reproducibility of existing results

Option C: Hybrid approach

• Use GENOMICON-Seq for capture + PCR simulation
• Pipe fragments to ReSeq for Illumina error profiles
• Pros: Best of both worlds (realistic duplicates + validated error model)
• Cons: Pipeline complexity

Technical Considerations

Input requirements

• Reference genome (we have: hg38/hg19)
• Probe BED file (need: kit-specific files for common platforms)
• Target regions (we have: MUC1 VNTR coordinates)

Output format

• Verify GENOMICON-Seq outputs standard BAM/FASTQ
• Check read naming conventions for duplicate tracking
• Ensure compatibility with downstream callers (VNtyper, adVNTR)

Success Criteria

• GENOMICON-Seq runs successfully with MUC1 VNTR region
• Duplicate family sizes show power-law distribution (not binary)
• Integration adds <20% runtime overhead vs current pipeline
• Clear documentation on when to use each simulator

Related Issues

• Investigation: Wessim duplicate generation model (sampling vs PCR amplification) #51: Wessim duplicate generation model investigation
• ReSeq seqToIllumina has 10k read limit (upstream bug) #41: ReSeq 10k read limit (may affect hybrid approach)

References

• GENOMICON-Seq: https://github.com/Gabaldonlab/GENOMICON-Seq
• PCR duplicate modeling: https://doi.org/10.1093/bioinformatics/btv024
• Wessim limitations: Investigation: Wessim duplicate generation model (sampling vs PCR amplification) #51

[berntpopp]

https://github.com/Rounge-lab/GENOMICON-Seq