Bayesian Inference of Dynamic Gene Regulatory Networks from Factorial Time-Course Experiments

• Activity Number: 16
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: International Society for Bayesian Analysis (ISBA)
• Author(s): Mayetri Gupta*+ and Joseph Wu and Louis Gerstenfeld
• Companies: University of Glasgow and Boston University School of Public Health and Boston University School of Medicine
• Keywords: mixture models ; Markov chain Monte Carlo ; clustering ; gene expression ; genomics

Abstract

Efficient microarray and high-throughput sequencing technologies developed in recent years allow experimenters to measure dynamic gene activity over time. Profiling dynamic transcriptional activity gives important insights into how genes respond over time to conditions such as exposure to pathogens, administration of a drug, or disease progression. More complex experimental designs, such as factorial time-course experiments, make it more challenging to answer questions of interest, such as how multiple experimental factors interact in their effects on gene expression over time. We propose a novel Bayesian approach that can simultaneously estimate longitudinal signals under such designs and assign genes into biologically meaningful clusters. Our methodology builds on the framework of hierarchical Bayesian longitudinal mixture models, and develops fast hybrid MCMC algorithms for model-fitting. This allows information about gene expression to be interpreted at all levels-longitudinal, factorial, and transcriptional. We applied our method on data collected from a longitudinal factorial study of fracture healing in mice, leading to several interesting and useful biological insights.