lapmix

lapmix applies a hierarchical Bayesian Laplace mixture model with empirical Bayes and likelihood-based hyperparameter estimation to identify differentially expressed genes in microarray experiments.

Key Features:

• Laplace Mixture Model: Implements a Laplace mixture model as an alternative to normal mixtures to accommodate long-tailed, variable, and skewed gene expression distributions.
• Hierarchical Bayesian Framework: Uses a hierarchical Bayesian approach to integrate prior information and model multi-level variation.
• Empirical Bayes Hyperparameter Estimation: Employs empirical Bayes methods to estimate hyperparameters from the data.
• Likelihood-Based Hyperparameter Estimation: Supports likelihood-based approaches for hyperparameter estimation, applicable to both Laplace and normal mixture models.
• Bidirectional Expression Detection: Detects both overexpression and underexpression of genes between conditions.
• Array Type Adaptability: Applicable to whole genome arrays and arrays with restricted coverage.
• Performance Characteristics: Simulation studies report some improvement in data fitting and comparable differential-expression detection relative to several existing statistical approaches.

Scientific Applications:

• Differential Expression Analysis: Identifies genes differentially expressed between two conditions, e.g., treatment versus control or wild type versus knockout.
• Biological and Disease Studies: Applied to microarray studies aimed at elucidating molecular mechanisms in complex disease traits and fundamental biological processes.

Methodology:

Uses a hierarchical Bayesian framework with a Laplace mixture model and employs empirical Bayes and likelihood-based methods to estimate hyperparameters.