Modelling Non‐homogeneous Dynamic Bayesian Networks with Piecewise Linear Regression Models

Abstract

In statistical genomics and systems biology non-homogeneous dynamic Bayesian networks (NH-DBNs) have become an important tool for learning regulatory networks and signalling pathways from post-genomic data, such as gene expression time series. This chapter gives an overview of various state-of-the-art NH-DBN models with a variety of features. All NH-DBNs, presented here, have in common that they are Bayesian models that combine linear regression with multiple changepoint processes. The NH-DBN models can be used for learning the network structures of time-varying regulatory processes from data, where the regulatory interactions are subject to temporal change. We conclude this chapter with an illustration of the methodology on two applications, related to morphogenesis in Drosophila and synthetic biology in yeast.