Biophysical Analysis for Implementing Genetic Associative Memory Using Hopfield Networks

To date, no reliable computing paradigms offer easy scaling of genetic networks with characteristics that are similar to natural systems. Recently, neuromorphic design has been investigated as a solution. Furthermore, the construction of biological systems with memory abilities having more than two-bits has not yet been accomplished. This work investigates the implementation of Hopfield networks in genetic circuits, which are capable of implementing associative memory and solving optimization problems. We developed a fundamental transformation to convert Hopfield networks to molecular and protein networks. Our approach transforms biochemical reactions to neural-like nodes having weights, biases, thresholds and activation functions. An energy function to such networks was derived and the convergence to a local minimum of the energy was proven. This work gives a proof of concept that neuro-inspired architecture can be implemented in genetic networks.