A Bayesian semiparametric factor analysis model for subtype identification.

Area of Science:

• Biomedical research
• Computational biology
• Genomics

Background:

• Disease subtype identification is crucial for understanding disease heterogeneity.
• Gene expression profiles are widely used for inferring disease subtypes.
• High dimensionality and gene correlation pose challenges for existing clustering methods.

Purpose of the Study:

• To introduce a novel Bayesian method, BCSub, for disease subtype identification.
• To address limitations of current clustering methods in high-dimensional gene expression data.
• To improve the identification of clinically relevant disease subtypes.

Main Methods:

• Developed a semiparametric Bayesian factor analysis model for dimension reduction.
• Utilized factor scores following a Dirichlet process mixture model for clustering.
• Evaluated the method through extensive simulation studies and application to real gene expression datasets.

Main Results:

• BCSub demonstrated improved performance compared to commonly used clustering methods in simulations.
• The method successfully identified clinically relevant subtypes in two gene expression datasets.
• BCSub effectively handles high dimensionality and gene correlation in gene expression data.

Conclusions:

• BCSub offers a robust and effective approach for disease subtype identification using gene expression profiles.
• The Bayesian framework and factor analysis model enhance clustering accuracy and clinical relevance.
• This method advances the field of computational biology for precision medicine.