Mutual information in the frequency domain for the study of biological systems

Many biological systems are comprised of multiple components that are interacting nonlinearly and producing multiple outputs of distinct frequency characteristics. Quantitative analysis of the observable outputs to identify the dependencies among components is imperative to increase the understanding of the underlying mechanism of the system. In this work, quantification of nonlinear dependencies in terms of mutual information between time series with respect to frequency characteristics is explored. A new model-free methodology is developed and tested on simulated data from coupled nonlinear systems. The results indicate that the proposed framework performs better than a conventional method for quantifying interactions. Application on real-world electrophysiological data from an emotional state assessment experiment reveals specific brain areas that are associated with levels of emotional responses.