Clustering Gene Expressions Using the Table Invitation Prior.

Area of Science:

• Computational Biology
• Statistical Genetics
• Bioinformatics

Background:

• Clustering gene expression data is crucial for understanding biological systems and disease subtypes.
• Traditional clustering methods often require manual parameter selection, which can be subjective and time-consuming.
• Bayesian nonparametric methods offer flexible approaches to clustering but can be complex to implement.

Purpose of the Study:

• To introduce and evaluate the Table Invitation Prior (TIP) for Bayesian nonparametric clustering of gene expression data.
• To demonstrate TIP's ability to automatically estimate the number of clusters without analyst intervention.
• To apply TIP for classifying cancer subtypes based on gene expression profiles.

Main Methods:

• Utilized the Table Invitation Prior (TIP), a Bayesian nonparametric clustering method.
• Employed pairwise subject distances as input for the TIP model.
• Estimated TIP hyperparameters using a univariate multiple change point detection algorithm.
• Implemented a Gibbs sampling algorithm in conjunction with a Normal-Inverse-Wishart likelihood.

Main Results:

• Successfully clustered 801 gene expression samples into distinct groups.
• Demonstrated automatic estimation of the number of clusters by TIP.
• The method effectively utilized pairwise distances for clustering gene expression data.
• The clustering was applied to samples from five different cancer types.

Conclusions:

• The Table Invitation Prior (TIP) provides an automated and effective method for clustering gene expression data.
• TIP's automatic hyperparameter estimation simplifies the clustering process for biological data analysis.
• This approach holds promise for subtype discovery and classification in complex diseases like cancer.