TopicNet: a framework for measuring transcriptional regulatory network change.

Area of Science:

• Genomics
• Systems Biology
• Computational Biology

Background:

• Large-scale chromatin immunoprecipitation sequencing (ChIP-seq) reveals dynamic changes in gene regulatory network connectivity across human cell states.
• Current regulatory networks are dense and noisy, hindering informative comparisons of transcription factor (TF) target gains and losses between cell states.

Purpose of the Study:

• To develop an abstracted, low-dimensional representation for understanding major features of gene regulatory network change.
• To quantify regulatory network rewiring and identify TFs with significant connectivity changes between cellular states, such as during oncogenesis.

Main Methods:

• Application of latent Dirichlet allocation (LDA) to extract functional topics from gene sets regulated by specific TFs.
• Definition of a 'rewiring score' to quantify regulatory network changes based on TF-specific topic alterations.
• Integration of gene expression data to define a 'topic activity score' for measuring topic activity in specific cellular states.

Main Results:

• TopicNet successfully extracts functional topics and quantifies TF rewiring across cellular states.
• The framework identifies TFs exhibiting significant network connectivity changes, relevant to processes like oncogenesis.
• Topic activity scores correlate with differential survival in various cancers, highlighting clinical relevance.

Conclusions:

• TopicNet provides a powerful framework for abstracting and analyzing complex gene regulatory network dynamics.
• The method offers novel insights into cellular state transitions and has potential applications in cancer research and personalized medicine.