We created artificial neural network type architecture with engineered bacteria to perform reversible and irreversible computation. This may work as new computing system for performing complex cellular computation.

Advancement of in-cell molecular computation requires multi-input-multi-output genetic logic devices. However, increased physical size, a higher number of molecular interactions, cross-talk, and complex systems level device chemistry limited the realization of such multi-input-multi-output devices in a single bacterial cell. Here, by adapting a circuit minimization and conjugated promoter engineering approach, we created the first 3-input-3-output logic function in a single bacterial cell. The circuit integrated three extracellular chemical signals as inputs and produced three different fluorescent proteins as outputs following the truth table of the circuit. First, we created a noncascaded 1-gate-3-input synthetic genetic AND gate in bacteria. We showed that the 3-input AND gate was digital in nature and mathematically predictable, two important characteristics, which were not reported for previous 3-input AND gates in bacteria. Our design consists of a 128 bp DNA scaffold, which conjugated various protein-binding sites in a single piece of DNA and worked as a hybrid promoter. The scaffold was a few times smaller than the similar 3-input synthetic genetic AND gate promoter reported. Integrating this AND gate with a new 2-input-2-output integrated circuit, which was also digital-like and predictive, we created a 3-input-3-output combinatorial logic circuit. This work demonstrated the integration of a 3-input AND gate in a larger circuit and a 3-input-3-output synthetic genetic circuit, both for the first time. The work has significance in molecular computation, biorobotics, DNA nanotechnology, and synthetic biology.

This work presented an application of genetic distributed computing, where an abstract computational problem was mapped on a complex truth table and solved using simple genetic circuits distributed among various cell populations. Maze generating and solving are challenging problems in mathematics and computing. Here, we mapped all the input-output matrices of a 2 × 2 mathematical maze on a 4-input-4-output truth table . The logic values of four chemical inputs determined the 16 different 2 × 2 maze problems on a chemical space. We created six multiinput synthetic genetic AND gates, which distributed among six cell populations and organized in a single layer. Those cell populations in a mixed culture worked as a computational solver, which solved the chemically generated maze problems by expressing or not expressing four different fluorescent proteins. The three available "solutions" were visualized by glowing bacteria, and for the 13 "no solution" cases, no bacteria glowed. Thus, our system not only solved the maze problems but also showed the number of solvable and unsolvable problems. This work may have significance in cellular computation and synthetic biology.