A neural network based modeling and validation approach for identifying gene regulatory networks

We present a comprehensive neural network based modeling and validation framework for inferring regulatory interactions from temporal gene expression data. We introduce gene set stochastic sampling and sensitivity analysis as two methods for identifying minimal regulatory elements of a target gene expression profile. We test the accuracy of these methods on a simulated dataset, and a biological animal model. A thorough computational approach is also presented to test the validity and robustness of the inferred regulations. We demonstrate that our modeling framework is able to accurately capture the majority of the known interactions in both the simulated and biological data.