Let it RE:IN: integrating experimental observations to predict pluripotency network behaviour

It is not unusual for biologists to sketch the interactions of genes or proteins in their system of interest in order to construct networks that describe the phenomenon that they study. Furthermore, from university lectures to graphical abstracts we use these sketches to demonstrate how we think the system works, often by using up and down arrows to describe dynamics. Increasingly, there are repositories or databases that we can mine to establish which interactions might occur in our data [for instance, STRING (Szklarczyk et al, 2010), Reactome (Joshi‐Tope et al, 2005) and KEGG (Kanehisa, 2000)]. However, these networks are constructed using pre‐existing knowledge; thus, they provide us with a graphical summary of the history of the nodes, but offer little insight into how these interactions respond to changes and shape complex behaviours. Although in the last years these networks have been utilised to make accurate predictions about how networks can be controlled (e.g., Cahan et al, 2014; D'Alessio et al, 2015; Rackham et al, 2016), these purely static networks cannot be directly probed to predict or confirm the dynamics of perturbations.

In this issue, Dunn et al (2019) tackle this problem in an unconventional way by applying their Reasoning Engine for Interaction Networks (RE:IN) approach, introduced earlier to establish a minimal regulatory network architecture required for ESCs self‐renewal (Dunn et al, …