Bayesian hierarchical model for transcriptional module discovery by jointly modeling gene expression and ChIP-chip...

Area of Science:

• Computational biology
• Genomics
• Systems biology

Background:

• Transcriptional modules (TMs) comprise co-regulated genes and their regulating transcription factors (TFs).
• High-throughput (HT) technologies like gene expression microarrays and Chromatin Immuno-Precipitation on Chip (ChIP-chip) provide genome-scale data on gene regulation.
• Integrating these complementary datasets for TM identification is a significant challenge in computational biomedicine.

Purpose of the Study:

• To develop and validate a novel probabilistic model for joint analysis of gene expression and ChIP-chip data.
• To identify and characterize transcriptional modules (TMs) more effectively.
• To improve the identification of gene regulatory relationships.

Main Methods:

• Developed a novel probabilistic model and computational procedure for joint analysis of gene expression and ChIP-chip data.
• Validated the method's performance against separate analyses and alternative joint approaches.
• Implemented the algorithm within the R package 'gimmR' for user accessibility.

Main Results:

• The new method yields TMs with improved functional coherence compared to analyzing data separately or using other joint methods.
• The algorithm successfully identifies novel regulatory relationships missed by ChIP-chip data alone.
• The computational procedure is user-friendly, comparable to hierarchical clustering, requiring minimal data transformation.

Conclusions:

• Joint analysis of ChIP-chip and expression data within a unified probabilistic framework is recommended for improved TM identification and regulatory relationship discovery.
• The developed computational procedure is a valuable tool for reconstructing transcriptional regulatory networks due to its statistical properties and ease of use.