Linear and nonlinear correlation estimators unveil undescribed taxa interactions in microbiome data

Area of Science:

• Microbiology
• Bioinformatics
• Statistical Modeling

Background:

• Human microbiomes involve complex microbial ecosystems with nonlinear interactions.
• Current methods lack the ability to quantify these nonlinear relationships effectively.
• Understanding these interactions is crucial for microbiome research.

Purpose of the Study:

• To develop a novel methodology for estimating linear and nonlinear microbial correlations.
• To introduce Sparse Estimation of Correlations among Microbiomes (SECOM) for microbiome data analysis.
• To address limitations in existing measures for quantifying microbial interdependencies.

Main Methods:

• Developed SECOM, a method for estimating sparse linear and nonlinear correlations among microbes.
• SECOM incorporates models for sample and taxon-specific biases.
• Statistical properties were validated analytically and through simulations.

Main Results:

• SECOM identified high correlations between skin microbiomes (forehead and palm), outperforming competing methods.
• The study characterized temporal changes in bacterial family correlations in infant gut microbiomes during the first year of life.
• SECOM revealed previously uncharacterized nonlinear relationships in microbiome data.

Conclusions:

• SECOM is a powerful tool for quantifying complex linear and nonlinear microbial interactions.
• The method provides novel insights into skin and infant gut microbiome dynamics.
• SECOM advances the field of microbiome analysis by enabling the study of intricate microbial relationships.