RioNorm2

RioNorm2 normalizes microbiome count data and performs differential abundance testing using a network-based size factor estimation and a two-stage zero-inflated mixture count regression model. It addresses zero-inflation, over-dispersion, and compositional bias in microbial community datasets.

Key Features:

• Network-Based Normalization: Identifies relatively invariant microbial species across samples to construct size factors for accurate normalization.
• Two-Stage Zero-Inflated Mixture Regression: Models zero-inflation and over-dispersion using a mixture count regression framework for differential abundance analysis.
• Flexible Dispersion Modeling: Accommodates species-specific dispersion characteristics to reduce false positives and improve detection of small effect sizes.
• Comparative Performance Validation: Demonstrates improved statistical power and robustness relative to DESeq, DESeq2, metagenomeSeq, RAIDA, Omnibus, and ANCOM in simulation studies.

Scientific Applications:

• Microbiome Differential Abundance Analysis: Identifies disease-associated or condition-specific microbial species in clinical and ecological microbiome studies.

Methodology:

RioNorm2 constructs normalization size factors from network-identified invariant species, then applies a two-stage zero-inflated mixture count regression model to account for zero-inflation and over-dispersion, enabling robust differential abundance testing across varying sample and library sizes.