SparseDOSSA 2

SparseDOSSA 2 simulates microbial abundance profiles to generate realistic synthetic microbiome datasets for evaluation and benchmarking of statistical methods.

Key Features:

• Statistical modeling: Represents marginal microbial feature abundances using a zero-inflated log-normal distribution to handle sparsity and zero-inflation.
• Parameterization from real data: Parameterizes models using real-world microbial community profiles to generate new realistic synthetic profiles with known structures.
• Covariance structure simulation: Simulates fully known covariance structures among synthetic features (taxa) and between features and phenotypes.
• Metadata association simulation: Simulates associations between microbial features and sample metadata and among microbial features.
• Spike-in of associations: Allows spiking-in true positive synthetic associations to create controlled signals for method evaluation.

Scientific Applications:

• Modeling human-associated microbial profiles: Models human-associated microbial population profiles for human microbiome research.
• Controlled synthetic community generation: Generates synthetic communities with specified population and ecological structures to test hypotheses and validate analytical methods.
• Benchmarking analysis methods: Produces datasets with known structures and spiked-in associations for benchmarking statistical and computational methods.
• Replication of experimental scenarios: Recapitulates end-to-end experimental scenarios, such as a mouse microbiome feeding experiment, for method validation.

Methodology:

Represents marginal abundances with a zero-inflated log-normal distribution; parameterizes model parameters from real microbial community profiles; simulates known covariance structures between taxa and between taxa and phenotypes; simulates associations with sample metadata; supports spiking-in true positive synthetic associations.