A Molecular Perceptron Based on DNA Circuit

Activation functions are critical to artificial neural networks; however, their implementation in DNA circuitry has remained a long-standing challenge, impeding the advancement of DNA circuits in artificial intelligence. This paper constructs a step function through idealized reactions and utilizes this function to implement a novel type of molecular perceptron. This perceptron exhibits promising linear classification behavior in simulation and is used to construct logic gates with semi-logical features including AND, OR, and NOT operations. These gates enable the implementation of complex logical circuits for a range of representative Boolean functions. Taking the XOR operation circuit as an example, the proposed molecular perceptron transforms DNA semi-logical circuits into DNA logical circuits. This method bridges the gap between molecular-level analog-to-digital conversion modules and digital-to-analog conversion modules, providing a possible design route toward the conversion interface between simulated signals within living organisms and digital signals outside the body.