LinDA

LinDA performs differential abundance analysis on compositional microbiome sequencing data by applying linear regression to centered log-ratio (CLR)-transformed features and maintaining asymptotic false discovery rate control.

Key Features:

• Centered log-ratio (CLR) transformation: Applies CLR transformation to mitigate compositional effects in sequencing-derived abundance data.
• Linear regression modeling: Uses linear regression models on CLR-transformed data for taxon-level differential abundance testing.
• Compositional bias correction: Corrects biases introduced by the compositional constraints of microbiome data.
• Asymptotic FDR control: Maintains asymptotic false discovery rate (FDR) control for identified differential features.
• Mixed-effect model extension: Can be extended to mixed-effect models to accommodate correlated microbiome data from longitudinal or hierarchical designs.
• Validation: Performance has been assessed using simulations and real-world examples.

Scientific Applications:

• Differential abundance testing: Identifying taxa with statistically significant differences across experimental conditions or groups in microbiome studies.
• Longitudinal and hierarchical study analysis: Detecting differential signals in correlated samples from longitudinal or nested experimental designs.
• Method validation and benchmarking: Using simulations and empirical datasets to evaluate differential abundance detection and FDR control.
• Microbiome–health associations: Investigating microbial differences related to health and disease in compositional sequencing data.

Methodology:

LinDA applies centered log-ratio (CLR) transformation to compositional abundance data, fits linear regression models on CLR-transformed features, implements asymptotic FDR control, and can be extended to mixed-effect models for correlated data; validation was performed via simulations and real-world examples.