gemerr

gemerr generates simulated next-generation sequencing reads using empirically derived, sequence-context-based error models to reproduce run- and technology-specific error profiles for analyses such as SNP calling.

Key Features:

• Empirical Error Modeling: Derives sequence-context-based error profiles from real sequencing runs, including modeling of indels and quality score variations.
• Format and Platform Support: Produces single- and paired-end reads compatible with SAM and FASTQ formats and reflects platforms such as Illumina and Roche/454.
• Versatile Simulation Capabilities: Simulates reads from one or more genomes or haplotype sets to support diverse experimental scenarios.
• Error Profile Comparison: Derives and compares error models from multiple sequencing runs to characterize run-to-run and technology-specific differences.
• Impact on Variant Calling: Demonstrates effects of error profiles on SNP-calling accuracy, including findings that VarScan retains accuracy primarily for SNPs with frequency greater than 3% and that optimal VarScan 'minimum average quality' settings vary with error model.
• Empirical Fragment and Quality Distributions: Generates reads reflecting observed fragment length and quality score distributions from actual sequencing experiments.

Scientific Applications:

• Deep sequencing evaluation: Assess how sequencing error profiles affect detection of low-frequency variants in deep sequencing experiments.
• Metagenomic studies: Simulate complex communities to evaluate how run-specific errors influence taxonomic and diversity analyses.
• Resequencing projects: Benchmark variant-calling pipelines and parameter settings under realistic error conditions.
• Variant-calling validation: Quantify the impact of different error models on SNP-calling accuracy and on parameters such as VarScan's 'minimum average quality'.
• Genetic diversity assessment: Explore how sequencing-run variability in errors alters estimates of genetic diversity and low-frequency SNP detection.

Methodology:

Uses empirical sequencing run data to derive context-specific error models; generates reads that reflect empirical fragment length and quality score distributions; analyzes how differing error profiles affect SNP-calling accuracy and parameter sensitivity (e.g., VarScan 'minimum average quality').