metaMix

metaMix applies a Bayesian mixture model to probabilistically infer species composition from metagenomic sequencing data, with emphasis on detecting low-abundance viral pathogens in clinical samples.

Key Features:

• Bayesian Mixture Model Framework: Employs a Bayesian mixture model to probabilistically infer sample composition and account for uncertainty from ambiguous short sequencing reads.
• Parallel Monte Carlo Markov Chains (MCMC): Uses parallel MCMC sampling to explore the species space and identify the most likely set of contributing species.
• Focus on Viral Pathogen Detection: Designed to detect low-abundance viral pathogens in clinical metagenomic samples where reference databases may be incomplete.
• Empirical Performance: Demonstrated improved accuracy in profiling communities composed of several related species in empirical evaluations.

Scientific Applications:

• Clinical metagenomics: Detection and profiling of infectious pathogens, including low-abundance viruses, from deep sequencing data.
• Complex community profiling: Resolving closely related species and mixed-species datasets in metagenomic samples.
• Research in microbiology, virology, and environmental science: Analysis of mixed-species datasets to characterize community composition and pathogen presence.

Methodology:

MetaMix implements a Bayesian mixture model with parallel Monte Carlo Markov Chain (MCMC) sampling to probabilistically assign short metagenomic sequencing reads to species.