A learning system based on signaling network in living cells

Abstract

A mathematical model of the learning system based on signaling networks in living cells has been proposed. It is well known that a living body has the ability to adapt to environment, and each cell consisting the body may have the same ability. According to various researches in biochemical field, it is expected that the sensors perceiving surroundings are membrane proteins, and the actuators in cells are actin filaments. But the controller for actuators and adaption mechanism of cells has not been understood. When an extracellular signaling molecule activates a protein located on the cell membrane, this protein triggers a biochemical chain in the cell. The network of chemical action is the system which determines the response of the cell to extracellular stimuli. This “signaling network” is expected to work as adaptive control system in living cells. A similar network known as a neural network has ability of learning and its mathematical model has been applied to various fields such as control, pattern recognition, etc. An actual neural network in living body uses electric signals, but the signaling network uses diffusion of various molecules to transmit information. In this paper, a mathematical model of the signaling network having different mechanism from neural networks has been proposed. The model has been applied to nonlinear mapping problem and it shows ability as a learning system. Moreover, a learning ability for time series has been introduced by time delay mechanism to the present mathematical model.