Applications of Granger causality to biological data

In computational biology, one often faces the problem of deriving the causal relationship among different elements such as genes, proteins, metabolites, neurons and so on, based upon multi-dimensional temporal data. In literature, there are several well-established reverse-engineering approaches to explore causal relationships in a dynamic network, such as ordinary differential equations (ODE), Bayesian networks, information theory and Granger Causality. To apply the four different approaches to the same problem, a key issue is to choose which approach is used to tackle the data, in particular when they give rise to contradictory results. In this thesis, I provided an answer by focusing on a systematic and computationally intensive comparison between the two common approaches which are dynamic Bayesian network inference and Granger causality. The comparison was carried out on both synthesized and experimental data. It is concluded that the dynamic Bayesian network inference performs better than the Granger causality approach, when the data size is short; otherwise the Granger causality approach is better. Since the Granger causality approach is able to detect weak interactions when the time series are long enough, I then focused on applying Granger causality approach on real experimental data both in the time and frequency domain and in local and global networks. For a small gene network, Granger causality outperformed all the other three approaches mentioned above. A global protein network of 812 proteins was reconstructed, using a novel approach. The obtained results fitted well with known experimental findings and predicted many experimentally testable results. In addition to interactions in the time domain, interactions in the frequency domain were also recovered. In addition to gene and protein data, Granger causality approach was also applied on Local Field Potential (LFP) data. Here we have combined multiarray electrophysiological recordings of local field potentials in both right inferior temporal (rIT) and left IT (lIT) and right anterior cingulate (rAC) cortices in sheep with Granger causality to investigate how anaesthesia alters processing during resting state and exposure to pictures of faces. Results from both the time and frequency domain analyses show that loss of consciousness during anaesthesia is associated with a reduction/disruption of feed forward open-loop cortico-cortical connections and a corresponding increase in shorter-distance closed loop ones