Advances to Bayesian network inference for generating causal networks from observational biological data

TLDR

Network inference algorithms identify putative causal interactions from observational data, and Bayesian network inference can capture diverse linear, nonlinear, combinatorial, and stochastic relationships across biological levels, but limited experimental data remain a challenge. The study uses simulations to improve a dynamic Bayesian network inference algorithm for limited biological data by testing various scoring metrics and search heuristics. The authors evaluate sampling intervals, discretization levels, and introduce a novel influence score that estimates interaction sign and magnitude in DBNs. Combining the influence score with moderate data interpolation reduces false positives and improves DBN recovery from limited data, and the source code and simulated data are available.

Abstract

Network inference algorithms are powerful computational tools for identifying putative causal interactions among variables from observational data. Bayesian network inference algorithms hold particular promise in that they can capture linear, non-linear, combinatorial, stochastic and other types of relationships among variables across multiple levels of biological organization. However, challenges remain when applying these algorithms to limited quantities of experimental data collected from biological systems. Here, we use a simulation approach to make advances in our dynamic Bayesian network (DBN) inference algorithm, especially in the context of limited quantities of biological data.We test a range of scoring metrics and search heuristics to find an effective algorithm configuration for evaluating our methodological advances. We also identify sampling intervals and levels of data discretization that allow the best recovery of the simulated networks. We develop a novel influence score for DBNs that attempts to estimate both the sign (activation or repression) and relative magnitude of interactions among variables. When faced with limited quantities of observational data, combining our influence score with moderate data interpolation reduces a significant portion of false positive interactions in the recovered networks. Together, our advances allow DBN inference algorithms to be more effective in recovering biological networks from experimentally collected data.Source code and simulated data are available upon request.http://www.jarvislab.net/Bioinformatics/BNAdvances/