PBSIM2

PBSIM2 simulates long-read sequencing datasets from a reference genome, modeling quality scores and non-uniform error patterns characteristic of PacBio and Oxford Nanopore technologies.

Key Features:

• Generative model for quality scores: Uses a hidden Markov Model (HMM) with factorized information criteria to capture and simulate non-uniform error characteristics of long-read sequencing.
• Efficiency and scalability: Generates simulated datasets in minutes at high coverage (e.g., 100x) for large references (~10 million base pairs) with runtime scaling proportionally with coverage depth and reference length.
• Memory requirements: Requires memory roughly equal to the length of the reference sequence plus a few megabytes.
• Validation and consistency: Produces simulated reads that closely mimic real long-read sequencing data for reliable downstream analyses.

Scientific Applications:

• Pipeline development and testing: Produces realistic datasets for developing and benchmarking bioinformatics pipelines for long-read data.
• Sequence assembly: Enables evaluation and optimization of assembly algorithms under long-read error profiles.
• Variant calling and other genomic analyses: Facilitates testing of variant calling and other analyses that require accurate modeling of read errors.

Methodology:

Simulates sequencing reads from a reference genome while modeling quality scores using a hidden Markov Model combined with factorized information criteria to represent observed non-uniform error patterns.