A model selection criterion for model-based clustering of annotated gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Interpreting gene clusters in gene expression data often requires external functional annotations.
• Existing methods may not fully leverage incomplete or partial gene annotation information.

Purpose of the Study:

• To propose a novel model selection criterion for gene expression data analysis.
• To enhance the interpretability of gene clusters by integrating external functional annotations.

Main Methods:

• Developed the integrated completed annotated likelihood (ICAL) criterion based on finite mixture models.
• Incorporated an entropy term into a penalized likelihood to measure concordance between gene clusters and annotations.
• Applied ICAL in conjunction with Gaussian mixture models.

Main Results:

• The ICAL criterion effectively selects the optimal number of clusters and clustering model.
• ICAL facilitates the choice of gene clustering models that are more interpretable with respect to known functional annotations.
• Demonstrated performance on both simulated and real RNA-seq gene expression data.

Conclusions:

• ICAL provides an efficient and interpretable approach for analyzing gene expression data.
• This method improves the biological relevance of gene cluster interpretation by utilizing functional annotations.
• ICAL is a valuable tool for researchers in genomics and bioinformatics.