Enhanced Feature Selection for Microbiome Data using FLORAL: Scalable Log-ratio Lasso Regression.

Area of Science:

• Microbiome research
• Computational biology
• Biostatistics

Background:

• High-throughput microbiome data analysis is crucial for identifying patient outcome biomarkers.
• Existing computational tools often fail to adequately address complex survival endpoints, longitudinal samples, and sequencing biases.
• Accurate microbial feature selection is essential for reliable biomarker discovery.

Approach:

• Introducing FLORAL, an open-source computational tool for scalable log-ratio lasso regression.
• FLORAL handles continuous, binary, time-to-event, and competing risk outcomes.
• The method incorporates longitudinal microbiome data as time-dependent covariates and employs a two-stage screening process for enhanced false-positive control.

Key Points:

• FLORAL demonstrates superior false-positive control compared to existing lasso-based methods.
• It offers improved sensitivity over differential abundance testing methods, especially for smaller sample sizes.
• FLORAL significantly enhances microbial feature selection by leveraging longitudinal data in survival analyses.

Conclusions:

• FLORAL provides a robust and scalable solution for microbiome-based biomarker discovery.
• The tool effectively integrates complex survival endpoints and longitudinal microbiome data.
• FLORAL represents a significant advancement in analyzing microbiome data for predicting patient outcomes.