Enzymatic Networks for Pattern Recognition: Basic Principles and Applications

模式识别作为机器学习的一个子主题，被定义为获取原始数据并根据数据的类别采取行动的行为。本研究探索了利用连接可调节的下游分子逻辑门网络分析多个激活传感器输出的可能性，并由此得到一个分类，即单个输出1（动作）或0（无动作）。基于核酸催化剂（脱氧核酶）的逻辑门接受并分析一个或多个寡核苷酸作为输入，并产生反应，如催化活性，通常通过荧光的增加来可视化。作为逻辑门输入的寡核苷酸可以被纳入基于寡核苷酸的识别区域（适体）。这种设计可用于构建基于脱氧核酶和适体的纯分子专家系统，该系统将与生理条件兼容，并且不使用电子元件或人工输入。本研究的最终目标是构建能够识别和纠正代谢状态变化的自主治疗分子装置。

Pattern recognition, as a subtopic of machine learning, is defined as the act of taking in raw data and acting based on the category of the data. This research explores the possibility to analyze outputs from multiple activated sensors using networks of downstream molecular logic gates with adjusted connections, and as a result obtain a classification, that is, a single output of 1 (Action) or 0 (No Action). Logic gates based on nucleic acid catalysts (deoxyribozymes) accept and analyze one or more oligonucleotides as inputs, and produce a response, such as catalytic activity, usually visualized by an increase in fluorescence. The oligonucleotides that serve as inputs for logic gates can be incorporated into oligonucleotide-based recognition regions (aptamers). Such design can be used to construct a purely molecular expert system based on deoxyribozymes and aptamers, which will be compatible with physiological conditions, and will not use electronic components, or human inputs. The eventual goal of this research is to construct autonomous therapeutic molecular devices that recognize and correct changes in metabolic states.