Identification of important regressor groups, subgroups and individuals via regularization methods: application to...

Area of Science:

• Microbiome research
• Statistical bioinformatics
• Host-microbe interactions

Background:

• Gut microbiota analysis is complex due to multiple taxonomic levels.
• Targeting specific microbial levels for health requires robust identification methods.
• Existing statistical approaches often fail to consider multiple taxonomy levels simultaneously.

Purpose of the Study:

• To develop a novel statistical method for identifying important microbial features across multiple taxonomic levels.
• To enable more precise targeting of gut microbiota interventions for health improvements.
• To provide a flexible and robust tool for microbiome data analysis.

Main Methods:

• Developed a new variable selection method using L1 and L2 regularizations.
• Implemented a data-adaptive, repeated cross-validation approach for parameter selection.
• Applied the method to analyze gut microbiota composition in relation to interventions and physiological status.

Main Results:

• The new method successfully identifies significant microbial features at multiple taxonomic levels.
• Simulation studies demonstrated superior performance compared to existing methods, with high detection rates and low error rates.
• Analysis of gut microbiota data revealed specific taxonomic levels most affected by interventions and physiological states.

Conclusions:

• The developed method offers a powerful approach for dissecting complex gut microbiota data.
• It facilitates the identification of crucial microbial targets for therapeutic interventions.
• The R package implementation ensures accessibility for researchers in the field.