Informative Structure Priors: Joint Learning of Dynamic Regulatory Networks From Multiple Types of Data

Abstract

We present a method for jointly learning dynamic models of transcriptional regulatory networks from gene expression data and transcription factor binding location data. Models are automatically learned using dynamic Bayesian network inference algorithms; joint learning is accomplished by incorporating evidence from gene expression data through the likelihood, and from transcription factor binding location data through the prior. We propose a new informative structure prior with two advantages. First, the prior incorporates evidence from location data probabilistically, allowing it to be weighed against evidence from expression data. Second, the prior takes on a factorable form that is computationally efficient when learning dynamic regulatory networks. Results obtained from both simulated and experimental data from the yeast cell cycle demonstrate that this joint learning algorithm can recover dynamic regulatory networks from multiple types of data that are more accurate than those recovered from each type of data in isolation.

A Bayes Random Fields Approach for Integrative Large-Scale Regulatory Network Analysis

“…Another reason lies in that less direct relationships exist between PPI and the transcriptional network, since the PPI data can only indicate potentials for interactions rather than presence of interactions in the process under analysis. This is also consistent with previous findings [1]. Nonetheless, the inclusion of a data set of low relevance and high noise reflects the robustness of the proposed algorithm, since the resultant network is neither biased to noise nor affected by the irrelevant information.…”

“…Such gene expression dynamics are important since they directly reveal the active components within the cell over time, thereby indicating certain interacting relationships. Regarding to how the time series data is used in network reconstruction, two types of methods, Dynamic Bayesian networks (DBNs) [1] and Graphical Gaussian models (GGMs) [11], account for a major part of current research.…”

“…Therefore, by integrating multiple data types, one can expect false positives to be reduced and disparities between different levels of the system to be identified. Further, integration helps to explain complex biological interaction on a deeper level than using a single data alone [1].…”

“…Very often, integrative systems are constrained to certain two or three types of data [1,3,4]. For example, protein-protein interaction data and gene expression data are used in constructing probabilistic network models [3].…”

Learning Gene Regulatory Networks via Globally Regularized Risk Minimization

“…Learning the structure of a Bayesian network from data generally requires one of two approaches to be followed: a score-based approach-where a heuristic search is performed through the space of causal network structures to identify the most likely structure explaining the data-and a constraint-based approach-where conditional independence tests are used to determine whether a direct causal relationship should be postulated between two variables. Many variants of these techniques have been applied to gene regulatory network learning, including search-based approaches [6][7][8], information-theoretic approaches [9], parametertying based approaches [10], and conventional dynamic Bayesian network learning approaches [11,12].…”