Conformation-Driven Molecular Computing

点击翻译按钮获取中文摘要

CONRAD, MichaelCCR-9610054Wayne State UniversityThe high recognition capacity of proteins and other biological macromolecules derives largely from shape-based interactions. The pictures of a lock-key or jigsaw-like fit is frequently used to describe this type of molecular level pattern recognition. The goal of this research project is to harness this powerful intrinsic form of pattern recognition to solve computational problems. The basic idea is captured in the self-assembly model of computing. Signal patterns impinging on a device are transduced to molecular conformations which then self-organize to form a higher level complex (or mosaic). Shape features common to different mosaics represent different families of input patterns. Enzymes that recognize these shape features control the output of the device. The solution of potentially difficult symbolic pattern recognition problems is converted in this way to a macromolecular self-organization process driven by free energy minimization. Devices of this type cannot be programmed in a conventional sense, because of the difficulty of predicting the structural and functional consequences of conformational interactions. However, they are well suited to molding through directed evolution (due to the high structure-function plasticity of biological macromolecules) and to adaptation through error feedback acting on the internal structure of the device.Chemical implementation of this scheme can operate on the basis of the conformational responses of single proteins to different patterns of milieu features. Shape-based enzyme recognition serves to transform input signal patterns (coded in chemical form) to output activity (detected spectroscopically). The particular protein being used for this purpose is mitochrondrial malate dehydrogenase, a readily available enzyme of carbohydrate metabolism, that could in the future easily be coupled to other enzymes to build up a repertoire of special purpose pattern processors that can be utilized as co-processors for a conventional machine linked into computational networks capable of performing complex recognition-action tasks.