A Novel Linear Machine Learning Method Based on DNA Hybridization Reaction Circuit

DNA hybridization reaction is a significant technology in the field of semi-synthetic biology and holds great potential for use in biological computation. In this study, we propose a novel machine learning model based on a DNA hybridization reaction circuit. This circuit comprises a computation training component, a test component, and a learning algorithm. Compared to conventional machine learning models based on semiconductors, the proposed machine learning model harnesses the power of DNA hybridization reaction, with the learning algorithm implemented based on the unique properties of DNA computation, enabling parallel computation for the acquisition of learning results. In contrast to existing machine learning models based on DNA circuits, our proposed model constitutes a complete synthetic biology computation system, and utilizes the "dual-rail" mechanism to achieve the DNA compilation of the learning algorithm, which allows the weights to be updated to negative values. The proposed machine learning model based on DNA hybridization reaction demonstrates the ability to predict and fit linear functions. As such, this study is expected to make significant contributions to the development of machine learning through DNA hybridization reaction circuits.