FLORAL

FLORAL performs scalable log-ratio lasso regression and microbial feature selection for continuous, binary, time-to-event, and competing risk outcomes, with compatibility for longitudinal microbiome data as time-dependent covariates.

Key Features:

• Compositional feature selection: Selects predictive microbial biomarkers from high-throughput microbiome data while accounting for the compositional nature of sequencing data.
• Diverse outcome support: Handles continuous, binary, time-to-event, and competing risk outcomes.
• Longitudinal data compatibility: Accommodates longitudinal microbiome data as time-dependent covariates.
• Two-stage screening process: Implements a two-stage screening approach for enhanced false-positive control.
• Taxa-specific sequencing bias handling: Accounts for taxa-specific sequencing biases in microbiome data.
• Zero-sum constraint optimization: Adapts the augmented Lagrangian algorithm for zero-sum constraint optimization in log-ratio regression.

Scientific Applications:

• Survival analysis: Identifies predictive microbial biomarkers for complex survival endpoints and competing risk outcomes in clinical cohorts such as allogeneic hematopoietic cell transplant recipients.
• Longitudinal microbiome analysis: Improves microbial feature selection by utilizing longitudinal microbiome sampling over baseline-only analysis.
• Biomarker discovery: Detects predictive microbial features associated with patient outcomes from high-throughput sequencing data, particularly in smaller sample size datasets.

Methodology:

FLORAL adapts the augmented Lagrangian algorithm for a zero-sum constraint optimization problem and enables a two-stage screening process for enhanced false-positive control.