Implementing an artificial neural network using a synthetic bacteria quorum sensing system

Implementing synthetic systems are essential to engineering living organisms to solve real-world applications. Although synthetic biology has been progressed for over a decade, building intelligent systems that can recognize patterns has not been well studied. In the present work, we use Quorum Sensing (QS), an intercellular communication mechanism, to build a system that can recognize and classify input patterns in engineered living cells. Using the genetic parts involved in QS, we implemented bacteria network to allow several input patterns of N-Acyl homoserine lactones (AHL) to be classified. The network consists of sender cells that take the AHL inputs to express signaling molecules and receiver cells that detect the signaling molecules to produce Green Fluorescence Proteins (GFP). The sender and receiver cells constitute a one-layer perceptron network. We mutated the PluxR promoter in the genetic circuit of the sender cells at its initial four nucleotides, so that the circuit can be activated by AHL at varying strength and release a varying amount of signaling molecules. The mutations of PluxR promoter in the sender circuits are effectively the weights of the perceptron network. By arranging the weights of the network, we built a genetic classifier for AHL input patterns. We tested the system using a 2-pixel by 2-pixel pattern set. Tests using more sophisticated patterns are undergoing. The project lays the foundation of building an intelligent system in living organisms via intercellular communications.