Function-switch in evolvable mechanical networks

Elastic mechanical networks provide a physical model for adaptable matter. Distant pairs of source and target nodes can be coupled similarly to allosteric interactions in proteins. Two networks tuned for in-phase and for out-of-phase responses are related to each other by a discrete set of "mutations" each of which adds or removes a bond in the network. We showed that the effects of these mutations are epistatic (that is, they do not simply add linearly), and studied the behavior along the evolutionary pathways between networks with opposite functions. We found a surprisingly large critical response threshold above which no viable pathways persist. At this critical threshold, these systems are exceedingly adaptable – the function-switch is caused by a single bond, whose alteration dramatically switches the system's response between two extraordinarily fit states. We explore the mechanical origins and sensitivity of these "jumper" mutations. In most cases, the mutations break up into two distinct classes suggesting that there is a memory preserved in the pathway of the two variants' common ancestor. We analyze how different network tuning algorithms influence evolvability. This study enhances our understanding of the mechanical basis underlying functional adaptability and highlights the surprising efficiency with which mechanical networks can be evolved.