Gene Network Inference Using Forward Backward Pairwise Granger Causality

Abstract

Discovery of temporal dependence is the basic idea for evaluating gene networks using Granger causality. However, with the advancement of technology, now we can analyze multiple genes simultaneously that result in high dimensional data. Recent studies suggest that more causal information can be retrieved if we reverse the time stamp of time series data along with standard time series data. Based on these findings, we are proposing a new method called Forward Backward Pair wise Granger Causality. The results how that our method can handle high dimensional data and can extract more causal information compared to the standard ordinary least squares method. We have performed a comparison of proposed and existing method using simulated data and then used the proposed method on real Hela cell data and mapped the 19 genes that are commonly present in cancer.