Measuring Information Transfers in Biological Networks


Background: Information talk is pervasive in biology, yet it remains highly controversial among theoretical biologists and philosophers of biology. Here, we propose to ground information talk in measures of causal influence in biological networks. Roughly, how informative a cause will be with respect to an effect in a given system will depend on the extent to which the cause influences this effect in the system.Observations: We analyze two measures of causal influence: (1) causal specificity, and (2) information flow. Causal specificity measures the amount of information that a biological system or an experimenter can transmit to an effect variableby manipulating a cause variable. Causal specificity thus measures the amount of potential control of the effect through the cause. Information flow measures the amount of information actually flowing from a causal variable to an effect variable in some actual system. Information flow thus measures the extent to which the cause actually explains the effect in the system. We exemplify the behavior of these two measures using simple examples of complex systems, i.e. systems including interactions. We show that both measures are left unaffected by confounding factors, enabling them to measure causal influence and not mere association. Both measures also do justice to interactions, being sensitive to relevant background factors which may or may not be controlled while performing the measure. Differences between the two measures exist as causal specificity is more flexible regarding the manipulations of the system that are considered when performing the measure.Conclusions: Both causal specificity and information flow are suitable measures to study causal influence in biological networks. Both measures enable us to formally ground information talk in biology. The work presented here can be applied to more complex biological networks, such as gene regulatory networks.

European Molecular Biology Organization (EMBO) Meeting