Universal Computation via Self-assembly of DNA: Some Theory and Experiments

Abstract

In this paper we examine the computational capabilities inherent in the hybridization of DNA molecules. First we consider theoretical models, and show that the self-assembly of oligonucleotides into linear duplex DNA can only generate sets of sequences equivalent to regular languages. If branched DNA is used for self-assembly of dendrimer structures, only sets of sequences equivalent to context-free languages can be achieved. In contrast, the self-assembly of double crossover molecules into two dimensional sheets or three dimensional solids is theoretically capable of universal computation. The proof relies on a very direct simulation of a universal class of cellular automata. In the second part of this paper, we present results from preliminary experiments which investigate the critical computational step in a two-dimensional self-assembly process.