A Dirichlet process mixture model for clustering longitudinal gene expression data.

Area of Science:

• Biomedical research
• Computational biology
• Statistical genetics

Background:

• Subgroup identification is crucial in biomedical research, often using gene expression profiles.
• Longitudinal gene expression data offers deeper insights into disease progression than baseline data alone.
• Existing clustering methods struggle to fully leverage longitudinal gene expression data.

Purpose of the Study:

• To introduce a novel Bayesian clustering method, BClustLonG, designed for longitudinal gene expression profiles.
• To improve patient subgroup identification accuracy and effectiveness using time-series gene expression data.
• To address limitations of current statistical methods in utilizing longitudinal data for clustering.

Main Methods:

• Developed BClustLonG, a Bayesian clustering approach using a linear mixed-effects model for gene trajectory analysis.
• Employed a factor analysis model for regression coefficients to handle gene correlations and high dimensionality.
• Utilized Dirichlet process prior for regression coefficient means to induce patient clustering.
• Validated the method through extensive simulation studies and application to a severely injured patient dataset.

Main Results:

• BClustLonG demonstrated superior performance compared to existing clustering methods in simulations.
• The model successfully identified clinically relevant subgroups in a dataset of severely injured patients.
• The approach effectively models gene expression trajectories and facilitates accurate patient stratification.

Conclusions:

• BClustLonG offers a powerful new tool for subgroup identification in biomedical research using longitudinal gene expression data.
• The method enhances understanding of disease progression and patient heterogeneity.
• This Bayesian approach provides a robust framework for analyzing complex, time-dependent biological data.