A better sequence-read simulator program for metagenomics.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Existing whole-genome shotgun sequence simulators often use predefined models, limiting their applicability.
• Current simulators are typically restricted to single-genome studies and do not support non-parametric read-length distributions or quality profiles derived from metagenomics data.

Purpose of the Study:

• To develop a novel program, BEAR (Better Emulation for Artificial Reads), for generating simulated metagenomic sequencing reads.
• To enable the creation of simulated data that accurately reflects empirically-derived non-parametric read-length distributions and quality profiles from metagenomics datasets.

Main Methods:

• BEAR employs a machine-learning approach to emulate sequencing reads.
• It automatically determines parameter settings from user-supplied data, requiring minimal input.
• The program derives run-specific error rates and extracts quality-error models from metagenomic data.

Main Results:

• BEAR generates reads with lengths and quality values closely matching empirically-derived distributions.
• It can emulate reads from various sequencing platforms, including Illumina, 454, and Ion Torrent, and is particularly useful for technologies lacking dedicated simulators, such as Ion Torrent.
• BEAR automates the generation of abundance data, a previously arduous task in metagenomic simulation.

Conclusions:

• BEAR provides a significant advantage over existing software for evaluating genomics data processing tools.
• Its ability to generate more realistic, technology-independent reads makes it highly valuable for metagenomics research.
• The program's features streamline the simulation process, particularly for complex metagenomic datasets.