The concept of reservoir computing stems from the use of recursive connections within neural networks to create a complex dynamical system. wikipedia 
New research finds ion gradients across membranes form an information network for rapid cellular decision-making independent of DNA. post
Modeling non-genetic
information dynamics in cells
using reservoir computing. pdf
Living cells are non-linear dynamic systems that must constantly adapt to their environment. Modeling intracellular information dynamics solely based on physical and chemical laws needs many approximations and can still fall short of capturing the complexity of subcellular processes.
Thus, we developed Cell-Reservoir, a quasi-physical reservoir computing framework that pairs a grid graph-based reservoir system with a model for cellular decision-making processes capable of learning directly from measurements.
The Cell-Reservoir model consists of randomly generated networks of conducting cytoskeletons capable of transmitting electrical signals from the cell membrane to the various organelles, which are considered in this model as the local centers of decision-making.
Computationally, the Cell-Reservoir assumes voltage and current as node properties and conductivity as edge properties, which is stored in a spatially correct location in a grid-graph data structure, reducing both space and time complexity. This grid-graph structure, which comes from the network physical requirement, differs from the traditional RC framework.
In addition, each node stores a physical signal (current) and a memory signal (exponential time averages of the physical signal), enabling our model to effectively have a short-term memory of environ- mental perturbations and to learn dynamical cell behavior.