Multimodal probabilistic generative models for time-course gene expression data and Gene Ontology (GO) tags

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Gene expression analysis is crucial for understanding biological processes.
• Gene Ontology (GO) tags provide functional context for genes.
• Integrating gene expression and GO data presents analytical challenges.

Purpose of the Study:

• To propose novel probabilistic generative models for joint analysis of gene expression and GO tags.
• To leverage the complementary information between gene expression patterns and functional annotations.
• To apply these models to time-course data for studying biological processes like angiogenesis and cell cycles.

Main Methods:

• Developed four probabilistic generative models for simultaneous gene expression and GO tag modeling.
• Applied models to three time-course datasets (angiogenesis, mitotic cell cycles).
• Performed joint clustering of genes and GO annotations, enabling analysis at different time scales.

Main Results:

• Models achieved joint clustering of genes and GO annotations.
• Demonstrated utility for de novo biological stage boundary estimation.
• Showed improved biological stage prediction accuracy when GO tag information was included.
• Different models offered varying granularities of gene grouping based on GO tags and temporal behavior.

Conclusions:

• Joint modeling of gene expression and GO tags offers a powerful framework for biological discovery.
• Incorporating GO annotations generally improves the performance of temporal gene expression analysis models.
• The proposed models provide new tools for understanding dynamic biological processes and inferring biological stage information.