Bayesian hierarchically weighted finite mixture models for samples of distributions

Area of Science:

• Biostatistics
• Genetics
• Computational Biology

Background:

• Finite Gaussian mixture models approximate unknown probability density functions.
• DNA repair studies often involve analyzing cell samples from multiple individuals, requiring models that account for individual variability.

Purpose of the Study:

• To propose a novel class of hierarchically weighted finite mixture models for DNA repair data.
• To allow for heterogeneity in individual responses and predictor effects while using common Gaussian basis distributions.

Main Methods:

• Developed a flexible hierarchical modeling framework where individual-specific densities are mixtures of common Gaussian bases.
• Modeled mixture weights to capture individual heterogeneity.
• Incorporated analysis of variance-type structures and latent factor representations for simplified inference on covariance structures. Developed posterior computation methods.

Main Results:

• The proposed model effectively handles heterogeneity in DNA repair data.
• The model facilitated the selection of genetic predictors associated with DNA damage and repair.
• Simplified inferences were achieved for non-Gaussian covariance structures.

Conclusions:

• Hierarchically weighted finite mixture models offer a robust and flexible approach for analyzing complex biological data, such as DNA repair studies.
• The model successfully identified genetic factors influencing DNA damage and repair mechanisms.
• This framework enhances our understanding of individual variability in DNA repair processes.